Well logging optical cable fault root cause analysis system and method fused with knowledge graph

Abstract

The invention discloses a logging optical cable fault root cause analysis system and method fused with a knowledge graph, and relates to the technical field of fault root cause analysis. According to the system, in order to solve the problem of low recognition accuracy of a synergistic effect of corresponding composite incentives in a logging optical cable fault root cause analysis process, a fault domain knowledge graph adaptive to a complex underground scene is formed by calling a historical knowledge graph of underground environment monitoring of a historical logging optical cable and supplementing an entity relationship; meanwhile, the entity relationship is updated in combination with underground dynamic working condition information, then underground real-time multi-dimensional operation data are obtained and mapped to corresponding entity nodes of the updated knowledge graph, collaborative state recognition is completed, and finally, collaborative state recognition is performed according to the collaborative state recognition result. According to the method, fault root cause positioning is realized in the fault domain knowledge graph, and a targeted fault repair and operation and maintenance optimization scheme is generated, so that accurate identification of composite fault causes and efficient positioning of fault root causes in a complex underground scene are realized, and the accuracy and timeliness of fault analysis are improved.

Description

Technical Field

[0001] This invention relates to the field of root cause analysis technology, and in particular to a root cause analysis system and method for well logging optical cable faults that integrates knowledge graphs. Background Technology

[0002] In the oil and gas energy industry, logging fiber optic cables are the core infrastructure for downhole borehole measurement and data transmission, and their operational status directly determines the safety level and production efficiency of downhole operations. Currently, the industry has formed a product matrix of logging fiber optic cables centered on sensor cables, covering temperature sensor cables, vibration sensor cables, and stress-strain sensor cables, etc., adaptable to the monitoring needs of different downhole scenarios. With the deep application of the Internet of Things and sensor monitoring technologies in the energy exploration field, the operational data acquisition system for these logging fiber optic cables has been gradually improved, enabling real-time acquisition of multi-dimensional operational indicators such as transmission loss, temperature tolerance, and mechanical stress, providing ample data support for fiber optic cable condition monitoring.

[0003] However, the downhole operating environment is complex. High temperature and pressure, formation displacement, chemical corrosion, and other factors can all cause fiber optic cable failures. Once the fiber optic cable experiences transmission interruption or signal abnormalities, it can not only lead to the loss of downhole measurement data but also delay mining operations and cause safety accidents. Therefore, accurate root cause analysis of logging fiber optic cable failures, rapid location of fault causes, and development of repair plans have become crucial for ensuring stable downhole operations. Knowledge graph technology, with its powerful multi-source data association and modeling capabilities, can integrate information such as fiber optic cable equipment parameters, historical failure cases, and downhole environmental data to construct a network linking faults and their causes, providing a new technical approach for root cause analysis.

[0004] Existing technologies first construct a knowledge graph for the field of well logging optical cables. By sorting out information such as optical cable equipment parameters, historical failure cases, and downhole environmental data, entities and relationships such as failure types, causes, and influencing factors are constructed. Then, real-time monitoring data such as optical cable transmission loss and mechanical stress are accessed and mapped to corresponding entity nodes in the knowledge graph. Subsequently, the graph's association retrieval capabilities are used to match similar failure root cause records, and causal inference is performed in combination with preset reasoning rules. Finally, hierarchical conclusions of failure root causes are output, and corresponding operation and maintenance repair solutions are associated to complete the entire root cause analysis process.

[0005] However, there is still room for improvement in terms of adaptability to complex downhole scenarios and the accuracy of multi-factor correlation analysis, specifically in the following aspects:

[0006] Because logging cables are exposed to complex downhole environments for extended periods, they are susceptible to progressive sheath degradation caused by formation fluid chemical corrosion. The complex failure scenarios resulting from the superposition of this latent degradation process and sudden mechanical stress changes place a greater demand on the correlation analysis of multi-dimensional operational data and the prioritization of contributing factors. Adapting to these complex scenarios requires additional multi-factor correlation algorithms, leading to significant delays in the correlation analysis of corresponding failure causes, often failing to accurately identify the synergistic effects of multiple contributing factors.

[0007] In addition, under the aforementioned complex dynamic conditions, if a combination of a sudden temperature rise and micro-displacement of the formation occurs simultaneously downhole, that is, the association link weights of the two types of causal entities, namely temperature exceeding the standard and formation displacement, in the knowledge graph are static preset values ​​and cannot be dynamically updated according to the synchronous effect intensity of the two types of factors in real-time conditions. However, the actual fault may be the result of the superposition of multiple factors, which may directly lead to the reasoning conclusions being biased towards a single causal factor, reducing the matching degree between the root cause of the fault and the actual working conditions, and thus causing a deviation between the root cause analysis conclusions and the actual fault causes, resulting in insufficient targeting of the operation and maintenance plan.

[0008] Under the aforementioned dynamic operating conditions, the superposition of problems such as the failure to adapt the entity association weights of the knowledge graph in real time and the solidification of the knowledge graph architecture leads to the problem that the knowledge graph is not accurate enough in identifying the synergistic effect of compound fault causes. This limits the accuracy and timeliness of the root cause analysis of well logging optical cable faults, which not only causes the ineffective consumption of operation and maintenance resources, but also delays the repair time due to fault location deviations, further exacerbating the safety risks of downhole operations. Summary of the Invention

[0009] To address the issue of low accuracy in identifying the synergistic effects of multiple fault causes using knowledge graphs in existing technologies, this invention provides a system and method for analyzing the root causes of optical cable faults in well logging that integrates knowledge graphs. The technical solution is as follows:

[0010] On the one hand, a root cause analysis system for logging fiber optic cable faults integrating knowledge graphs is provided. This system includes: a dynamic knowledge graph construction and update module, used to retrieve historical knowledge graphs of downhole environmental monitoring corresponding to historical logging fiber optic cables, and supplement entity relationships to obtain a fault domain knowledge graph for fault analysis in complex downhole scenarios, while updating entity relationships in conjunction with dynamic downhole operating information; a composite fault cause association analysis module, used to acquire multi-dimensional operational data from real-time downhole monitoring, and map it to the entity nodes corresponding to the fault domain knowledge graph after entity relationship updates, and perform collaborative state identification; and a root cause localization and operation and maintenance scheme output module, used to locate the root cause of logging fiber optic cable faults in the fault domain knowledge graph based on the collaborative state identification results sent by the composite fault cause association analysis module, and generate fault repair and operation and maintenance optimization schemes.

[0011] On the other hand, a method for root cause analysis of logging fiber optic cable faults integrating knowledge graphs is provided. This method includes: retrieving historical knowledge graphs of downhole environmental monitoring corresponding to historical logging fiber optic cables and supplementing entity relationships to obtain a fault domain knowledge graph for fault analysis in complex downhole scenarios, while updating entity relationships in conjunction with downhole dynamic operating condition information; acquiring multi-dimensional operational data from real-time downhole monitoring and mapping it to the entity nodes corresponding to the fault domain knowledge graph after the entity relationship update, and performing collaborative state identification; based on the results of collaborative state identification, locating the root cause of logging fiber optic cable faults in the fault domain knowledge graph, and generating fault repair and operation and maintenance optimization schemes.

[0012] The beneficial effects of the technical solutions provided by the embodiments of the present invention include at least the following:

[0013] 1. This invention first constructs a fault domain knowledge graph adapted to complex downhole scenarios by retrieving historical knowledge graphs from well logging fiber optic cable downhole environmental monitoring and supplementing entity relationships. Simultaneously, it updates entity relationships in real time with dynamic operating condition information, effectively solving the problem of inference conclusions biased towards a single cause under complex operating conditions such as sudden temperature rises and formation micro-displacements, thus improving the matching degree between root causes and actual operating conditions. Secondly, it acquires real-time multi-dimensional downhole operating data, maps it to corresponding entity nodes in the updated fault domain knowledge graph, and completes collaborative state identification. This step accurately captures the synergistic effect of causes in complex fault scenarios involving the superposition of latent corrosion degradation and instantaneous mechanical stress mutations, improving the accuracy of multi-factor correlation analysis. Finally, based on the collaborative state identification results, it completes fault root cause localization in the fault domain knowledge graph and generates targeted operation and maintenance plans, improving the accuracy and timeliness of fault analysis, reducing ineffective consumption of operation and maintenance resources, avoiding delays in repair due to fault location errors, and effectively reducing downhole operation safety risks.

[0014] 2. By vectorizing structured data from multi-source heterogeneous data through character-level encoding, a fault feature vector set is obtained. Feature vectors suitable for complex operating conditions are then selected through sliding window sampling. Incremental learning is used to identify newly added fault entity types and convert them into entity nodes. After aligning with historical graph nodes, the correlation is verified. Finally, probabilistic inference is used to complete weight allocation and construct a fault domain knowledge graph. This achieves effective integration of multi-source data and accurate inclusion of new entities, improving the graph's adaptability to complex operating conditions. Entity relationship updates are divided into two categories: basic and instantaneous operating conditions. Basic operating condition updates calculate the dynamic change gradient of operating condition data and basic offset, mapping and filtering the basic data that induces faults. Initial weight values ​​are output and calibrated through probabilistic inference, synchronously updating the graph links to achieve dynamic adaptation between basic operating condition changes and entity relationships, avoiding misjudgments of causes caused by static weights. The instantaneous operating condition update calculates the real-time rate of change, extracts fluctuation parameters, maps and filters key instantaneous data, outputs multi-inducing factor synergy coefficients through linkage verification, and obtains the distribution density by fitting the weight contribution degree with the Gaussian kernel density function. High correlation relationships are marked and the spectrum is updated. By capturing the synergistic effect of composite instantaneous operating conditions, the accuracy of multi-factor correlation analysis is improved, providing dynamic and reliable spectrum support for subsequent synergistic state identification and ensuring the accuracy of fault root cause analysis.

[0015] 3. The collaborative state recognition in this invention is mainly divided into two categories: entity node linkage state recognition and fault feature coupling state recognition. In the entity node linkage state recognition stage: firstly, multi-dimensional dynamic monitoring data is acquired and mapped to the updated fault domain knowledge graph, matching entity node linkage states and determining collaborative state verification labels; through data interval judgment within a preset monitoring window and verification of synchronous change trends of dual working condition parameters, three types of labels are divided: no linkage anomaly, linkage anomaly, and severe linkage anomaly, and targeted composite cause analysis and root cause localization are performed. This process accurately adapts to the scenario of superimposed basic and instantaneous working conditions, realizing the hierarchical identification of fault causes under composite working conditions, and improving the targeting and efficiency of cause investigation. In the fault feature coupling state identification stage: multi-dimensional fault characterization data is acquired and mapped to the fault feature entity nodes of the graph matching. Based on the number of matching nodes, fluctuation time sequence deviation, and cross-link area of ​​entity relationships, four types of labels are defined: no coupling anomaly, binary coupling anomaly, multi-source coupling anomaly, and single feature isolated anomaly. Cause analysis and root cause localization are carried out according to the classification. When multi-source coupling anomaly occurs, full-link tracing and historical database matching verification are triggered. This process effectively solves the problem of difficulty in judging the coupling state in the scenario of cross-type fault cause interaction, realizes accurate classification of compound causes, avoids misjudgment of single causes, and improves the accuracy of root cause tracing through historical data matching, providing a reliable basis for subsequent accurate root cause localization.

[0016] 4. First, acquire composite cause localization data, including cause-related time-series response data reflecting the time-series response characteristics of multi-dimensional operational data, and link matching records showing the matching status of entity nodes and fault cause nodes. Extract instantaneous delay feature values ​​through intrinsic mode decomposition, calculate the proportion of missing cause dimensions, and fuse the two to obtain a comprehensive quantitative value of root cause localization effectiveness interference for judgment. If the comprehensive quantitative value meets the standard, after completing root cause localization, divide the impact range hierarchy according to the map node degree parameter, and generate a targeted repair and maintenance plan; if it does not meet the standard, return to supplement data collection to improve the link records until the standard is met. This process effectively avoids root cause localization deviation and improves localization accuracy by quantifying localization interference through multi-dimensional data fusion; by dividing the impact range hierarchy, it ensures that the maintenance plan accurately matches the degree of fault impact, avoiding waste of maintenance resources; the supplementary data collection and improvement mechanism further ensures the reliability of the localization results, providing strong support for the efficient handling of well logging fiber optic cable faults. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the structure of a well logging optical cable fault root cause analysis system that integrates knowledge graphs, provided in an embodiment of the present invention.

[0019] Figure 2 This is a flowchart of entity node linkage status recognition provided in an embodiment of the present invention;

[0020] Figure 3 A flowchart of a well logging optical cable fault root cause analysis method integrating knowledge graphs is provided in an embodiment of the present invention;

[0021] Figure 4 A flowchart for fault feature coupling state identification provided in an embodiment of the present invention;

[0022] Figure 5 A logical flowchart of a knowledge graph for root cause analysis of well logging optical cable faults provided in an embodiment of the present invention;

[0023] Figure 6 This is a schematic diagram of the structure of the fault domain knowledge graph provided in an embodiment of the present invention;

[0024] Figure 7 This is a schematic diagram of the structure of the knowledge graph corresponding to the fault domain in the collaborative state recognition process provided in an embodiment of the present invention. Detailed Implementation

[0025] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0026] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0027] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0028] Example 1

[0029] This invention provides a well logging optical cable fault root cause analysis system that integrates knowledge graphs, such as... Figure 1 The diagram shown illustrates the structure of a well logging fiber optic cable fault root cause analysis system that integrates knowledge graphs. This system may include:

[0030] The dynamic knowledge graph construction and update module is used to retrieve the historical knowledge graph of the downhole environment monitoring corresponding to the historical logging optical cable, and supplement the entity relationship to obtain the fault domain knowledge graph for fault analysis in complex downhole scenarios. At the same time, it combines the dynamic working condition information of the downhole to update the entity relationship, thereby improving the knowledge graph's ability to identify the collaborative state of compound fault causes.

[0031] The specific process for supplementing entity relationships includes: First, acquiring multi-dimensional data reflecting the current downhole environment and the corresponding logging fiber optic cable's operational status and historical maintenance history, denoted as multi-source heterogeneous data. Using a character-level encoding algorithm, the data fields of each structured data (structured operation monitoring data and semi-structured equipment parameters) are converted into fixed-length numerical codes. The numerical codes are then matched and regularized, and the results are combined. Finally, Z-score standardization is used to eliminate dimensional differences, outputting a fault feature vector set with unified dimensions and clear characteristics. Multi-source heterogeneous data typically includes three main categories: structured operation monitoring data, semi-structured equipment parameters, and unstructured maintenance log data.

[0032] Specifically, the structured operation monitoring data includes: distributed fiber optic sensors laid along the entire length of the logging cable core to collect real-time transmission loss values; a miniature stress-strain acquisition instrument fixed to the outer wall of the cable sheath to acquire sheath stress-strain values ​​at a preset sampling frequency; formation micro-displacement monitoring probes buried in the surrounding formations along the cable laying path to monitor formation micro-displacement data; and corrosive medium concentration monitoring sensors installed on the outer walls of the cable junction boxes and easily corroded sections to collect the concentration of surrounding corrosive media.

[0033] Semi-structured equipment parameters: extracted from the optical cable factory technical manual and historical downhole equipment management database, including formatted data such as optical cable model matching parameters and historical fault association tags; Unstructured operation and maintenance log data: entered on-site by downhole operators through mobile terminals (such as explosion-proof smart inspection mobile phones and industrial-grade rugged tablets), usually involving downhole operators' inspection records, descriptions of equipment abnormalities, emergency handling operation texts, and other unformatted character data.

[0034] Subsequently, a sliding time window was used to divide the sampling interval according to the fluctuation cycle of the complex working condition of the logging fiber optic cable. The time window length was set to 5-10 minutes and the step size to 2 minutes. Structured data that meets the current complex working condition of the logging fiber optic cable within the sampling interval was retained, and vector transformation was performed using a character-level encoding algorithm to obtain fault feature vectors that conform to the complex working condition of the logging fiber optic cable. Based on the incremental learning algorithm, the types of newly added fault entities under the current complex working condition were identified, such as fiber microbending entities induced by the superposition of instantaneous pressure impact and formation micro-displacement, and sheath cracking entities induced by the coupling of mixed corrosive media and temperature gradient. The attribute features such as the inducing conditions, fault manifestations, and impact range of the newly added fault entities were standardized into the data format preset by the knowledge graph, and assigned unique identifiers and classification attribute labels. The data was then analyzed using knowledge graph entity parsing tools (such as Python). The RDFLib library transforms standardized attribute data, unique identifiers, and classification tags into identifiable entity nodes in a knowledge graph. It then uses an attribute similarity matching algorithm to align these identifiable entity nodes with fault cause nodes in the historical knowledge graph. The library summarizes the number of alignments between entity nodes and fault cause nodes and compares it with the total number of identifiable entity nodes. If the number of alignments does not match the total number of entity nodes, it prompts a designated user to verify the association of the unaligned nodes. For example, it prompts a user to manually verify whether there are any missing entity nodes that couple mixed corrosive media with temperature gradients.

[0035] Conversely, a pre-defined probabilistic inference model (such as the NoisyOr model) is used to complete relationship probability verification and weight allocation. The NoisyOr model is constructed as follows: initializing the causal association parameters corresponding to historical fault cause nodes and historical entity nodes, setting a noise factor to quantify the probability deviation of a single cause independently triggering a fault, and generating a conditional probability table by iteratively optimizing the parameters through maximum likelihood estimation to complete the model construction. The NoisyOr model is chosen because well logging fiber optic cable faults are often caused by multiple independent or superimposed causes, which can accurately adapt to this scenario. It also has fewer parameters, is computationally efficient, and meets the needs of real-time downhole analysis. The input of the NoisyOr model is the fault feature vector and the structured data corresponding to the aligned entity nodes. The output is the association probability value between each fault cause node and the corresponding entity node. Based on the output association probability value, entity nodes whose association probability value is greater than the corresponding pre-defined probability value (set based on the confidence statistics of historical fault cause nodes and corresponding entity nodes) are input into the historical knowledge graph to obtain a fault domain knowledge graph adapted to complex downhole conditions. Entity relationships are updated in conjunction with dynamic downhole conditions information, including:

[0036] First, through the gradient equation Calculate the basic operating data of the well (such as temperature, pressure, and concentration of corrosive media) within the preset monitoring cycle. The dynamic gradient within the preset monitoring period, where t1 represents the start time of the preset monitoring period and t2 represents the end time of the preset monitoring period. and These correspond to the real-time collected values ​​of the basic operating condition data at times t1 and t2, respectively; the historical operating condition baseline values ​​corresponding to each basic operating condition data are retrieved, and the dynamic gradient is calculated using the gradient equation. The gradient values ​​of the historical operating conditions in the same dimension are respectively compared with the baseline gradient values. Perform the difference calculation using the following formula: The basic offset of each basic working condition data is obtained. This is to clarify the degree and direction of deviation of data from historical stable states under various operating conditions, such as the dynamic change gradient of corrosive medium concentration within the monitoring period. The concentration was 0.8 mg / (L·h), and the historical baseline gradient value was... The corresponding baseline offset is 0.2 mg / (L·h). The concentration was 0.6 mg / (L·h), and the direction was an increase in the concentration of the corrosive medium.

[0037] Then, the acquired basic offset is input into the fault domain knowledge graph. Entity nodes with basic offsets greater than those set by historical fault critical condition samples are retained. These are the basic condition data corresponding to the nodes, typically used to quantify the impact of basic condition offsets on logging fiber optic cable faults. The larger the basic offset, the greater the impact. Finally, the selected basic condition data is input into the NoisyOr model. The attribute feature library of entity nodes in the fault domain knowledge graph is called, and the association probability between the condition offset and each entity node is calculated using the built-in conditional probability table of the NoisyOr model. Initial values ​​of entity relationship weights are output and mapped to the [0,1] interval to eliminate dimensional differences, thereby obtaining dynamic entity relationship weight values. These values ​​are then synchronized to the corresponding entity relationship links in the fault domain knowledge graph to complete the update of entity relationships associated with basic conditions and perform collaborative state identification. The entity relationship links represent the causal association paths between fault causes, basic conditions, and fault features inherent in the fault domain knowledge graph.

[0038] The composite fault cause correlation analysis module is used to acquire multi-dimensional operational data from downhole real-time monitoring and map it to the corresponding entity nodes in the fault domain knowledge graph after the entity relationship is updated, so as to perform collaborative state identification and improve the accuracy and timeliness of fault cause correlation analysis.

[0039] Because the long-term deviation of the downhole basic operating conditions and the sudden fluctuation of instantaneous operating conditions have a superimposed effect, the causes of logging fiber optic cable failures are easily complicated. Therefore, when the logging fiber optic cable is based on the coordinated state corresponding to the multi-dimensional operating data, which corresponds to the superimposed fluctuation state of the basic operating conditions and instantaneous operating conditions, the corresponding coordinated state identification under the superimposed scenario of basic operating condition deviation and instantaneous operating condition fluctuation is performed, that is, entity node linkage state identification, such as... Figure 2 The flowchart shown is for recognizing the linkage status of entity nodes. Figure 2 The first preset condition refers to: the multi-dimensional dynamic monitoring data at each time within the preset monitoring window are all within the corresponding preset allowable range; the multi-dimensional dynamic monitoring data at each time within the preset monitoring window have a single data point whose fluctuation range exceeds the corresponding allowable fluctuation range, and there is another single data point whose fluctuation range within the preset monitoring window shows a synchronous increasing or decreasing trend with the previous single data point that exceeds the allowable fluctuation range.

[0040] The specific process for identifying the linkage status of entity nodes is as follows: Acquire multi-dimensional dynamic monitoring data (such as instantaneous pressure peaks, sudden temperature change rates, and sudden increases in corrosive medium concentration) to quantify the correlation between instantaneous operating condition deviations and fault induction. Map this data to the updated fault domain knowledge graph according to data type. Based on the precise matching mapping rules of data type-entity attribute labels in the fault domain knowledge graph, match the linkage status of the corresponding entity nodes, determine the collaborative status verification label, and perform composite cause analysis.

[0041] If the multi-dimensional dynamic monitoring data at each moment within a preset monitoring window (e.g., 10 minutes) are all within the corresponding preset allowable range, then the collaborative status verification label is determined to be without linkage anomaly, indicating no current fault cause. The preset allowable range is usually set by the average value of historical multi-dimensional dynamic monitoring data within historical monitoring windows. If, within the preset monitoring window, there is a single data point whose fluctuation amplitude exceeds the corresponding allowable fluctuation amplitude (e.g., instantaneous pressure fluctuation amplitude exceeds 0.5 MPa), and another single data point (e.g., temperature change rate) whose fluctuation amplitude within the preset monitoring window shows a synchronous increasing or decreasing trend with the previous single data point exceeding the allowable fluctuation amplitude, that is: If the correlation coefficient of the change values ​​corresponding to the fluctuation amplitudes of the two is greater than or equal to 0.8 (based on the summation and averaging of the correlation coefficients of historical change values ​​obtained under the same historical working conditions), then the collaborative status verification label is determined to be an abnormal linkage, indicating the existence of a composite fault cause. The correlation coefficient of the change values ​​represents the ratio of the product of the covariance and the corresponding standard deviation of the fluctuation amplitudes of the two within the preset monitoring window. Otherwise, the collaborative status verification label is determined to be a severe linkage abnormality, and a full-process cause tracing process is immediately triggered to conduct a full-link association investigation of all entity nodes in the fault domain knowledge graph (including fault cause nodes and entity nodes corresponding to basic working condition data). After the above composite cause analysis is completed, the root cause of the logging fiber optic cable fault is located.

[0042] The root cause localization and operation and maintenance solution output module is used to locate the root cause of well logging optical cable faults in the fault domain knowledge graph based on the collaborative status identification results sent by the composite fault cause association analysis module, and generate targeted fault repair and operation and maintenance optimization solutions, thereby reducing the safety risks of downhole operations.

[0043] The root cause localization of fiber optic cable faults in well logging includes: First, acquiring composite cause localization data of entity nodes in the corresponding fault domain knowledge graph after composite cause analysis. This composite cause localization data includes cause-related time-series response data and entity node link matching records. The cause-related time-series response data represents the feature set of multi-dimensional operational data corresponding to each composite cause experiencing response delays in the monitoring time-series dimension. Typically, multi-dimensional operational data is collected at a preset sampling frequency (e.g., 1 time / second) using downhole distributed monitoring equipment (such as fiber optic sensors or stress acquisition instruments), synchronously recording the collection timestamps of each data point. Then, based on the composite cause analysis results, multi-dimensional operational data sequences associated with each composite cause are selected, and the phases of each data point in the sequence are extracted. The delay duration at the baseline response time is summarized to form a response delay feature set; multi-dimensional operational data includes one or two of the following: multi-dimensional dynamic monitoring data and multi-dimensional fault characterization data; entity node link matching records are used to reflect the matching status between entity nodes and fault cause nodes in the fault domain knowledge graph. Typically, when multi-dimensional operational data is mapped to the fault domain knowledge graph, the node matching interface of the knowledge graph is called to automatically compare the entity attributes corresponding to the operational data with the label features of the fault cause nodes, record the successfully matched node pairs and the unmatched entity nodes, and supplement the historical matching results of node associations by combining historical matching logs, ultimately forming entity node link matching records between entity nodes and fault cause nodes.

[0044] Then, for the acquired causal time-series response data, an intrinsic mode decomposition algorithm is used to iteratively select intrinsic mode functions that satisfy the condition that the number of extreme points and zero-crossing points are equal or differ by 1. The formula is as follows: This formula separates the intrinsic mode function components of each order. In the formula, i represents the order of the intrinsic mode function, i=2,3,4…L, L represents the total order of the intrinsic mode functions, and j represents the index of the first i-1 order intrinsic mode functions. Represents the i-th order intrinsic mode function component. This represents the time-series response data associated with the original triggers. This represents the components of the first j-th eigenmode function. This represents the cumulative sum of the first i-1 order intrinsic mode function components. Based on the separated intrinsic mode function components of each order, the phase information of each component at the corresponding time is extracted. The phase difference between adjacent time points is calculated to obtain the instantaneous delay. Then, the instantaneous delay sequence of multi-dimensional operating data corresponding to each composite cause is extracted. Each value in the sequence is taken as the instantaneous delay feature value to obtain the instantaneous delay feature value of multi-dimensional operating data corresponding to each composite cause, thereby realizing the quantification of the lag characteristics of the time-series response of different causes. The number of misaligned nodes between entity nodes and fault cause nodes in the entity node link matching record is summarized. The ratio of the number of misaligned nodes to the total number of entity nodes in the entity node link matching record is processed to obtain the proportion of missing cause dimensions.

[0045] Finally, the peak values ​​of the instantaneous delay feature values ​​corresponding to different composite causes are obtained, and the maximum difference between any two peak values ​​is calculated and denoted as the peak difference. The proportion of missing cause dimensions is mapped to the [0,1] interval using the Min-Max method to obtain the normalized value. Calculate the covariance between the peak difference and the normalized value, where k represents the number of samples of the instantaneous delay feature value, n represents the total number of samples of the instantaneous delay feature value, and A k This represents the peak difference corresponding to the k-th sample. B represents the sample mean of the peak difference. k This represents the normalized value corresponding to the k-th sample. The sample mean of the normalized values. This is used to quantify the combined interference of temporal feature differences in causes and missing knowledge graph information on root cause localization, i.e., the synergistic effect of their biases at the sample level: if the biases are in the same direction (increasing / decreasing simultaneously), the covariance value will be amplified, reflecting the superposition effect of interference; if the biases are in opposite directions, the covariance value will be reduced, reflecting the offsetting effect of interference. Finally, a comprehensive quantitative value is obtained to quantify the degree of interference in the effectiveness of root cause localization, and root cause localization is determined accordingly.

[0046] When the comprehensive quantitative value is greater than the preset comprehensive quantitative value obtained by summing and averaging the historical comprehensive quantitative values ​​obtained in the historical logging fiber optic cable fault root cause localization process, it indicates that the logging fiber optic cable fault root cause localization is invalid, and prompts the preset personnel to manually check the logging fiber optic cable fault root cause localization process.

[0047] When the comprehensive quantification value is not greater than the preset comprehensive quantification value, it indicates that the root cause localization of the well logging fiber optic cable fault is effective, and the root cause localization of the well logging fiber optic cable fault is completed. Then, through the association links of entity nodes in the fault domain knowledge graph, the instantaneous delay feature value corresponding to the comprehensive quantification value and the proportion of missing cause dimensions are input into the labeled fault root cause type library, thereby locating the specific root cause type, such as the combined cause of fiber microbending and sheath corrosion. The number of entity nodes corresponding to the root cause type in the fault domain knowledge graph is obtained, and hierarchical division is performed based on the number of nodes. For example, if the number of nodes is ≥8, it is determined to be a first-level influence level. (Involving multiple optical cable segments and surrounding geological conditions), the corresponding fault repair and maintenance optimization plan is to prioritize shutting down the well to isolate the faulty segment, simultaneously replace the damaged optical cable core and sheath, and add geological micro-displacement monitoring points; if: 5 ≤ the number of nodes < 8, it is determined to be a level 2 impact (involving multiple components of a single optical cable segment), the plan is to replace the optical cable junction box and the sheath of the corroded section under pressure, and add a real-time monitoring threshold warning for optical fiber transmission loss; if the number of nodes < 5, it is determined to be a level 3 impact (single component failure), the plan is to supplement the anti-corrosion coating of the sheath, and carry out a special inspection once a week to ensure the relevance and operability of the plan.

[0048] The specific values ​​were set through statistical analysis of the historical impact range during the root cause localization process of well logging fiber optic cable faults. For example, reviewing the root cause localization analysis process of well logging fiber optic cable faults over the past 5 years, it was found that when the number of physical nodes corresponding to the root cause is ≥8, the fault affects 3 or more fiber optic cable segments and is accompanied by surrounding working conditions such as abnormal formation stress and excessive downhole temperature and humidity, resulting in a wide impact range and high difficulty in handling. When the number of nodes is between 5 and 7, the fault is concentrated in multiple core components such as the splice box, sheath, and fiber core of a single fiber optic cable segment, requiring targeted replacement of multiple components. When the number of nodes is <5, the fault is limited to a single component (such as a single splice box sealing failure or a single sheath corrosion). Therefore, 5 and 8 were taken as the key critical values ​​for the impact level classification, so that the level determination results are highly consistent with the actual degree of harm of downhole fiber optic cable faults.

[0049] This invention provides a method for root cause analysis of well logging optical cable faults that integrates knowledge graphs, such as... Figure 3 The flowchart shown is a method for root cause analysis of well logging fiber optic cable faults that integrates knowledge graphs. The processing flow of this method may include the following steps:

[0050] Historical knowledge graphs of downhole environmental monitoring corresponding to historical logging optical cables are retrieved and entity relationships are supplemented to obtain a fault domain knowledge graph for fault analysis in complex downhole scenarios. Entity relationships are updated by combining this with dynamic downhole operating information. Multi-dimensional operational data from real-time downhole monitoring are acquired and mapped to the corresponding entity nodes in the updated fault domain knowledge graph for collaborative state identification. Based on the collaborative state identification results sent by the composite fault cause association analysis module, the root cause of the logging optical cable fault is located in the fault domain knowledge graph, generating a fault repair and maintenance optimization plan.

[0051] In Example 1, the collaborative design of three modules—dynamic knowledge graph construction and updating, composite fault cause correlation analysis, and root cause localization and operation and maintenance solution output—provides the following benefits: First, by incrementally learning to supplement newly added fault entities and combining dynamic operating condition updates to entity nodes, the fault domain knowledge graph becomes more adaptable to complex downhole overlapping operating conditions, improving its adaptability to composite fault scenarios. Second, the innovative adoption of a collaborative identification mechanism for basic operating condition offsets and instantaneous operating condition fluctuations, combined with correlation coefficients to quantify data linkage trends, effectively distinguishes between occasional fluctuations and collaborative fault-causing fluctuations, improving the accuracy and timeliness of composite fault cause identification and solving the problem of difficult fault localization caused by multiple overlapping downhole causes. Third, by extracting temporal delay features through inherent mode decomposition and fusing the proportion of missing cause dimensions to determine the effectiveness of root cause localization, the accuracy of root cause localization is ensured. At the same time, the scope of influence is divided based on the node degree parameter of the knowledge graph, and targeted operation and maintenance plans are generated to shorten the fault investigation and repair cycle, reduce the safety risks of downhole operations, and thus effectively improve the stability and reliability of the root cause analysis results of faults in the downhole environment where the extreme logging optical cable is located.

[0052] Example 1 describes entity relationship updates based on various downhole basic operating condition data. Similarly, based on Example 1, Example 2 is added to describe entity relationship updates based on various downhole instantaneous operating condition data.

[0053] Example 2

[0054] The entity relationship update based on downhole instantaneous operating condition data is performed as follows: First, the real-time change rate of downhole temperature, pressure, corrosive medium concentration, and other instantaneous operating condition data within a preset monitoring period is calculated using gradient equations. After calculating the real-time change rate, a sliding window method is used to fit the trend of the real-time change rate, and extreme points in the fitted curve are extracted. The difference between extreme points is used as the dynamic fluctuation amount, and the slope of the tangent line of the fitted curve is used as the fluctuation trend slope. This accurately describes the fluctuation amplitude and increase / decrease of the operating condition, thereby obtaining the dynamic fluctuation amount of each instantaneous operating condition data and the fluctuation trend slope reflecting the increase / decrease of the fluctuation. Next, the obtained dynamic fluctuation amount and fluctuation trend slope are synchronously input into the fault domain knowledge graph for instantaneous operating condition feature mapping. By comparing the preset operating condition feature thresholds and association rules in the knowledge graph, instantaneous operating condition data that can quantify the impact of instantaneous operating condition offset on logging fiber optic cable faults are selected.

[0055] Subsequently, the selected instantaneous operating condition data is input into the NoisyOr model. The NoisyOr model performs linkage verification on the correlation between composite fault causes, analyzes the impact of different instantaneous operating condition factors on optical cable faults, and outputs the multi-cause collaborative influence coefficient of the corresponding entity nodes in the fault domain knowledge graph. On this basis, the weight contribution of each entity relationship link is fitted and calculated by combining the preset Gaussian kernel density function. The weight ratio within the unit correlation interval is solved by integral operation, thereby obtaining the entity relationship weight distribution density. This density can intuitively reflect the contribution intensity of different entity relationships to the fault. Finally, entity relationships with entity relationship weight distribution density greater than the corresponding reference value (this reference value is set based on historical fault data statistical analysis and is used to distinguish high / low correlation entity relationships) are marked with high correlation priority, and these marked entity relationships are synchronously updated to the corresponding entity relationship links in the fault domain knowledge graph, completing the entity relationship update of instantaneous operating condition association, and providing accurate knowledge graph support for subsequent collaborative state identification of composite fault causes.

[0056] In Example 2, the fluctuation characteristics of instantaneous operating conditions are accurately captured by sliding window fitting and extreme point extraction, improving the effective utilization rate of instantaneous operating condition data. The NoisyOr model is used to realize the linkage verification of compound fault causes, and the entity relationship weight distribution density is fitted by Gaussian kernel density function, which can more accurately characterize the synergistic impact of multiple instantaneous operating conditions on optical cable faults, making the entity relationships of the fault domain knowledge graph more consistent with the complex downhole scenario. By prioritizing and dynamically updating highly correlated entity relationships, the adaptability of the knowledge graph to instantaneous operating condition changes is strengthened, effectively improving the timeliness and accuracy of fault analysis and reducing the risk of misjudgment of downhole optical cable faults.

[0057] Example 1 describes the identification of the linkage status of physical nodes under the superposition of basic working condition deviation and instantaneous working condition fluctuation. Similarly, if the downhole logging optical cable fault is not caused by the superposition of a single working condition factor, but by the cross-action of different types of fault causes such as mechanical damage, material aging, and electromagnetic interference, then based on Example 1, Example 3 is added to identify the collaborative status under the cross-action scenario of cross-type fault causes, which is called fault feature coupling status identification.

[0058] Example 3

[0059] like Figure 4The flowchart shown illustrates the process for identifying the coupling state of fault features. The specific steps are as follows: First, acquire multi-dimensional fault characterization data to quantify the cross-influence of various fault features, such as optical cable transmission attenuation, sheath deformation, and electromagnetic interference intensity. Then, map these data to the updated fault domain knowledge graph. Based on the precise matching rules of data type-entity attribute tags in the fault domain knowledge graph, match the corresponding fault feature entity nodes, determine the coupling state verification label, and perform composite cause analysis.

[0060] If only a single fault feature entity node is matched, and the fluctuation range of the corresponding fault characterization data exceeds the corresponding allowable fluctuation range within the preset monitoring window, then the coupling state verification label is determined to be an uncoupled anomaly with no compound cause. The fluctuation range of the corresponding fault characterization data represents the difference between the maximum and minimum values ​​of the fault characterization data within the preset monitoring window. The allowable fluctuation range is set by statistically analyzing the upper limit of the fluctuation range of the fault characterization data during the risk-free operation period of the same type of single fault in the past.

[0061] It is important to understand that this judgment logic focuses on an independent anomaly corresponding to a single fault feature entity node, without involving the correlation between multiple fault features. The root cause of the anomaly only points to the excessive fluctuation range of the fault characterization data corresponding to the logging fiber optic cable, which is a simple fault type caused by a single factor and does not involve coupled anomalies. Therefore, there is no need to investigate the composite causes.

[0062] If two types of fault feature entity nodes are matched, and the fluctuation time sequence deviation of the two types of nodes within the preset monitoring window is not greater than the preset allowable fluctuation time sequence deviation, then the coupling state verification label is determined to be a binary coupling anomaly, indicating the existence of a composite cause. The fluctuation time sequence deviation represents the time difference between the peak fluctuations in the fault characterization data corresponding to the two types of fault feature entity nodes. The preset allowable fluctuation time sequence deviation is set by summing and averaging the results of the linkage response times of the two types of fault features during the analysis of historical binary coupling fault composite cause analysis.

[0063] It is important to understand that the fluctuation timing deviation here is only a quantitative indicator for determining the correlation between the two types of fault characteristics. It points to the abnormal state of the logging optical cable itself, rather than the abnormality of the delay itself. The fluctuation timing deviation of the two types of nodes is not greater than the preset allowable fluctuation timing deviation, indicating that the two are caused by the linkage of composite factors related to the optical cable, such as the synergistic effect of optical cable sheath corrosion and optical fiber microbending, rather than a single optical cable fault or an independent delay problem.

[0064] If three or more fault feature entity nodes are matched, and the cross-link area of ​​the entity relationship links of each node in the fault domain knowledge graph is not less than the preset allowable coupling link area, then the coupling status verification label is determined to be multi-source coupling anomaly, with complex composite causes, and the full-link cause tracing mechanism is triggered. The cross-link area represents the area of ​​the topological region formed by the links of the three or more fault feature entity nodes that are interconnected in the fault domain knowledge graph. The larger the area, the higher the degree of association between the nodes. The preset allowable coupling link area is set by summing and averaging the topological areas of the corresponding entity relationship links in the historical multi-source coupling fault composite cause analysis process.

[0065] Except for the above situations (i.e.) Figure 4 If the second preset condition is not met, it is judged as a single feature isolated anomaly. This is because such scenarios do not meet the judgment conditions corresponding to uncoupled anomalies, binary coupled anomalies, and multi-source coupled anomalies. It is neither a single fault feature exceeding the threshold fluctuation, nor two or more fault features meeting the linkage timing / topology association requirements. It only manifests as a small fluctuation or non-associated fluctuation of a certain fault feature entity node. There is no superposition of compound causes. Therefore, it is not necessary to trigger full-link tracing. It is only necessary to trace the root cause for a single fault feature entity node. After the compound cause analysis, the root cause of the well logging optical cable fault is located.

[0066] To trace the root cause of a single fault-characteristic entity node, the following steps are taken: Based on the attribute tags and associated link data of the corresponding entity node, retrieve the associated operating condition monitoring link (e.g., the fiber core loss-junction box sealing link corresponding to transmission attenuation), and obtain the real-time data of each node in the corresponding operating condition monitoring link. Pre-assigned personnel then compare this data with historical abnormal data nodes from previous root cause tracing processes, and combine this with corresponding historical fault records to further determine the single root cause. For example, when tracing the entity node corresponding to transmission attenuation, if any data point in the humidity data of the junction box sealing node exceeds the corresponding allowable value (including the allowable values ​​corresponding to temperature and humidity), the corresponding root cause is junction box sealing failure. The allowable values ​​for temperature and humidity are based on historical temperature and humidity monitoring data of similar junction boxes under fault-free operation, and are statistically analyzed to set the corresponding upper limit. When tracing the sheath deformation degree, if the formation stress node data exceeds the formation stress safety value set by pre-assigned personnel based on the actual logging fiber optic cable safety requirements, the corresponding root cause is formation displacement compressing the fiber optic cable.

[0067] In Example 3, by accurately matching multi-dimensional fault characterization data (optical cable transmission attenuation, sheath deformation, etc.) with the updated fault domain knowledge graph, and combining the graded judgment of three types of quantitative indicators—fluctuation amplitude, fluctuation timing deviation, and cross-link area—the comprehensiveness of tracing the causes of complex faults is ensured while improving the processing efficiency of single anomalies. This provides accurate pre-positioning support for the root cause location of well logging optical cable faults, effectively reduces fault investigation costs, and improves the pertinence of operation and maintenance response.

[0068] It should be added that, such as Figure 5 The flowchart shown illustrates the logical flow of the knowledge graph used for root cause analysis of well logging fiber optic cable faults. First, two types of core meta-paths are extracted from the knowledge graph: one type is the component topology meta-path between the fiber core, repeater, photoelectric converter, and wellhead terminal, used to characterize the connection relationships of the fiber optic cable hardware components; the other type is the fault-root cause association meta-path between fiber core breakage, repeater power supply anomaly, excessive interface attenuation, and downhole pressure, used to associate fault phenomena with potential causes. Then, the Node2vec algorithm is used to perform random walks on these two types of meta-paths. By controlling the depth and breadth of the walks, sampling paths covering component topology and fault-root cause associations are generated, ensuring that the paths fully reflect the structural features of the knowledge graph. Finally, the generated sampling paths are input into the Skip-gram model. Utilizing the model's vector mapping capability, the well logging fiber optic cable component topology feature matrix and the fault-root cause association feature matrix are obtained, respectively. These two matrices transform the unstructured relationships of the knowledge graph into structured vector representations, providing a computable feature foundation for subsequent fault root cause matching and cause association analysis.

[0069] like Figure 6 The diagram shows the structure of the fault domain knowledge graph. This graph, centered on entities related to well logging fiber optic cable faults, constructs a multi-level network encompassing various nodes including fiber optic cable components, fault characteristics, and operational condition causes. Fault characteristic nodes contain key data tags such as fiber optic cable transmission attenuation and sheath deformation, while operational condition cause nodes are associated with parameters such as downhole temperature and formation stress. One-way arrows indicate causal relationships between nodes, clearly presenting the correlation links between operational condition changes, fault characteristics, and fault types, providing a fundamental structural support for fault analysis.

[0070] like Figure 7 The diagram shown illustrates the structure of the fault domain knowledge graph during the collaborative state recognition process. This graph... Figure 6 On the basic framework, new red bidirectional arrows have been added. These red bidirectional arrows represent the interactive relationships between different fault characteristic entity nodes, such as the linkage between optical cable transmission attenuation and electromagnetic interference intensity nodes, and sheath deformation degree and ground stress nodes. They intuitively present the coupling effect scenario of multiple fault characteristics in collaborative state identification, and provide a visual basis for subsequent quantitative index judgment and composite cause analysis.

[0071] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0072] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0073] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0074] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0075] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0076] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A well logging optical cable fault root cause analysis system integrating knowledge graphs, characterized in that, The system includes: The dynamic knowledge graph construction and update module is used to retrieve the historical knowledge graph of the downhole environment monitoring corresponding to the historical logging optical cable, and supplement the entity relationship to obtain the fault domain knowledge graph for fault analysis in complex downhole scenarios. At the same time, it combines the dynamic working condition information of the downhole to update the entity relationship. The composite fault cause association analysis module is used to acquire multi-dimensional operational data from downhole real-time monitoring and map it to the entity nodes corresponding to the fault domain knowledge graph after the entity relationship is updated, so as to perform collaborative state identification. The root cause localization and operation and maintenance solution output module is used to locate the root cause of well logging optical cable faults in the fault domain knowledge graph based on the collaborative status identification results sent by the composite fault cause association analysis module, and generate fault repair and operation and maintenance optimization solutions.

2. The well logging optical cable fault root cause analysis system integrating knowledge graph as described in claim 1, characterized in that, The specific process for supplementing entity relationships includes: Multi-dimensional data reflecting the current downhole environment and the corresponding logging fiber optic cable's operating status and historical maintenance are acquired and denoted as multi-source heterogeneous data. The structured data within is vectorized through character-level encoding to obtain a fault feature vector set. A sliding time window is used to classify and sample the acquired fault feature vector set in order to filter out fault feature vectors that conform to the complex working conditions of logging optical cables. At the same time, an incremental learning algorithm is used to identify the types of new fault entities under the current complex working conditions. The newly added fault entity types are converted into entity nodes and aligned with the fault cause nodes in the historical knowledge graph. The system summarizes the number of node alignments between entity nodes and fault-causing nodes, and compares it with the total number of corresponding entity nodes. If the number of node alignments is inconsistent with the total number of corresponding entity nodes, the system prompts the designated personnel to verify the association relationship of the unaligned nodes. Conversely, the relationship probability verification and weight allocation are completed by combining the preset probabilistic reasoning model to obtain the fault domain knowledge graph, and the entity relationship is updated by combining the dynamic working condition information of the well.

3. The well logging optical cable fault root cause analysis system integrating knowledge graph as described in claim 1, characterized in that, The entity relationship update based on downhole dynamic operating condition information includes: The gradient equation is used to calculate the dynamic change gradient of each basic working condition data in the well within the preset monitoring period. The difference is calculated between the gradient equation and the corresponding historical working condition benchmark value to obtain the basic offset of each basic working condition data. The acquired basic offset is input into the fault domain knowledge graph for feature mapping in order to filter out the basic operating condition data used to quantify the impact of basic operating condition offset on logging fiber optic cable faults. The selected basic working condition data is input into the preset probabilistic reasoning model for preliminary verification of the fault cause. The degree of correlation between entity nodes and basic working condition offset in the fault domain knowledge graph is output and recorded as the initial value of entity relationship weight. The initial values ​​of entity relationship weights are normalized and calibrated to obtain dynamically adapted entity relationship weight values, which are then synchronized to the entity relationship links corresponding to the fault domain knowledge graph to complete the update of entity relationships associated with basic working conditions and perform collaborative state identification.

4. The well logging optical cable fault root cause analysis system integrating knowledge graph as described in claim 1, characterized in that, The entity relationship update based on downhole dynamic operating condition information includes: The real-time change rate of downhole instantaneous working condition data within a preset monitoring period is calculated by gradient equation, and sliding window trend fitting and extreme point extraction are performed to obtain the dynamic fluctuation amount and fluctuation trend slope of each instantaneous working condition data. The acquired dynamic fluctuation amount and fluctuation trend slope are synchronously input into the fault domain knowledge graph for instantaneous working condition feature mapping, so as to filter out the instantaneous working condition data used to quantify the impact of instantaneous working condition offset on logging fiber optic cable faults. The selected instantaneous operating condition data is input into the preset probabilistic reasoning model to perform linkage verification of composite fault causes, and the multi-cause synergistic influence coefficient of the corresponding entity node in the fault domain knowledge graph is output. Based on the multi-inducing synergistic influence coefficient, the weight contribution of each entity relationship link is fitted and calculated by combining the preset Gaussian kernel density function. The weight ratio within the unit association interval is obtained through integral operation, and then the entity relationship weight distribution density is obtained. Entities with a weight distribution density greater than the corresponding reference value are marked as having high relevance and priority. The corresponding entities are then synchronized to the entity relationship links in the fault domain knowledge graph to complete the instantaneous working condition-related entity relationship update and perform collaborative state identification.

5. The well logging optical cable fault root cause analysis system fused with knowledge graph as described in claim 3 or 4, characterized in that, The process of performing collaborative state identification includes: Identification of the linkage status of entity nodes under the superposition of basic working condition deviation and instantaneous working condition fluctuation, or identification of the coupling status of fault features under the cross-effect of cross-type fault causes.

6. The well logging optical cable fault root cause analysis system integrating knowledge graph as described in claim 5, characterized in that, The specific process for identifying the linkage status of the entity nodes is as follows: Multi-dimensional dynamic monitoring data is acquired to quantify the correlation between instantaneous operating condition deviation and fault induction. This data is then mapped to the fault domain knowledge graph after entity relationship updates. The linkage status of the corresponding entity nodes is obtained through mapping, and collaborative status verification labels are determined for composite cause analysis. If the multi-dimensional dynamic monitoring data at each time within the preset monitoring window are all within the corresponding preset allowable range, then the collaborative status verification label is determined to be without linkage abnormality and without fault cause. If, within the preset monitoring window, the multi-dimensional dynamic monitoring data at each moment shows that the fluctuation range of a single data point exceeds the corresponding allowable fluctuation range, and another single data point shows that the fluctuation range of the single data point exceeds the allowable fluctuation range, and the synchronous change trend of the two data points meets the expected synchronization requirements, then the collaborative status verification tag is determined to be abnormal, indicating a fault cause. Otherwise, the collaborative status verification label is determined to be a serious linkage anomaly, and a prompt is made to trigger the full-process cause tracing process to conduct a full-link investigation of all entity nodes in the fault domain knowledge graph; After analyzing the combined causes, the root cause of the logging fiber optic cable fault was located.

7. The well logging optical cable fault root cause analysis system integrating knowledge graph as described in claim 5, characterized in that, The specific process for identifying the coupled state of fault features is as follows: Multi-dimensional fault characterization data is acquired to quantify the degree of cross-influence of various fault features, and mapped one by one to the fault domain knowledge graph after the entity relationship is updated to obtain the corresponding fault feature entity nodes. The coupling state verification label is determined, and composite cause analysis is performed. If only a single fault feature entity node is mapped, and the fluctuation range of the corresponding fault characterization data exceeds the corresponding allowable fluctuation range within the preset monitoring window, then the coupling state verification label is determined to be an uncoupled anomaly with no compound cause. If two types of fault characteristic entity nodes are mapped, and the fluctuation time sequence deviation of the two types of nodes within the preset monitoring window is not greater than the preset allowable fluctuation time sequence deviation, then the coupling state verification label is determined to be a binary coupling anomaly, indicating the existence of a composite cause. If the mapping yields three or more types of fault characteristic entity nodes, and the cross-link area of ​​the entity relationship links of each node in the fault domain knowledge graph is not less than the preset allowable coupling link area, then the coupling status verification label is determined to be a multi-source coupling anomaly, indicating the existence of complex composite causes, and the full-link cause tracing mechanism is triggered. Except for the above situations, it is determined to be a single feature isolated anomaly. Only the matched fault feature entity node is traced to its corresponding single root cause, and the full-link cause tracing mechanism is not triggered. After analyzing the combined causes, the root cause of the logging fiber optic cable fault was located.

8. The well logging optical cable fault root cause analysis system fused with knowledge graph as described in claim 6 or 7, characterized in that, The process of locating the root cause of optical cable faults in well logging includes: After analyzing the composite causes, obtain the composite cause location data of entity nodes in the corresponding fault domain knowledge graph. The composite cause location data includes cause-related time-series response data and entity node link matching records. The acquired cause-related time-series response data is subjected to intrinsic mode decomposition to obtain the instantaneous delay feature values ​​of multi-dimensional running data corresponding to each composite cause. At the same time, the number of misaligned nodes between entity nodes and fault cause nodes in the entity node link matching record is counted. The ratio of the number of misaligned nodes to the total number of entity nodes in the entity node link matching record is processed to obtain the proportion of missing cause dimensions. The peak difference of each instantaneous delay feature value and the normalized value of the proportion of missing cause dimensions are fused by covariance to obtain a comprehensive quantitative value for quantifying the interference degree of the effectiveness of root cause localization, and then root cause localization is determined.

9. The well logging optical cable fault root cause analysis system integrating knowledge graph as described in claim 8, characterized in that, The specific process for determining the root cause is as follows: If the comprehensive quantification value is not greater than the preset comprehensive quantification value, it indicates that the root cause location of the well logging optical cable fault is effective, the root cause location of the well logging optical cable fault is completed, and based on the root cause type located, the node degree parameter of the entity node in the fault domain knowledge graph is obtained, and the hierarchical division of the root cause influence range is carried out, thereby generating a targeted fault repair and operation and maintenance optimization plan. If the comprehensive quantification value is greater than the preset comprehensive quantification value, it indicates that the root cause location of the logging fiber optic cable fault is invalid, and the system returns to the composite fault cause correlation analysis module to re-collect multi-dimensional operating data, supplement and improve the entity node link matching records, until the re-acquired comprehensive quantification value is not greater than the preset comprehensive quantification value, and the root cause location of the logging fiber optic cable fault is completed.

10. A method for root cause analysis of well logging fiber optic cable faults integrating knowledge graphs, applied to the root cause analysis system for well logging fiber optic cable faults integrating knowledge graphs as described in any one of claims 1-9, characterized in that, Includes the following steps: Historical knowledge graphs of downhole environmental monitoring corresponding to historical logging optical cables are retrieved and entity relationships are supplemented to obtain a fault domain knowledge graph for fault analysis in complex downhole scenarios. At the same time, entity relationships are updated in combination with dynamic downhole operating information. Acquire multi-dimensional operational data from downhole real-time monitoring and map it to the corresponding entity nodes in the fault domain knowledge graph after entity relationship updates for collaborative status identification; Based on the results of collaborative state recognition, the root cause of well logging fiber optic cable faults is located in the fault domain knowledge graph, and fault repair and operation and maintenance optimization solutions are generated.

Images