An AI-based device operation and maintenance knowledge graph construction method and system

Abstract

The application relates to the technical field of knowledge graph construction, in particular to an AI-based device operation and maintenance knowledge graph construction method and system. The method comprises the following steps: obtaining historical operation and maintenance data of each operation and maintenance object, performing knowledge extraction to obtain each triple data, analyzing the text word vector similarity and semantic coherence degree between different triple data to obtain the operation and maintenance feature coherence degree of each triple data, combining the local outlier feature of the operation and maintenance feature coherence of each triple data to obtain the semantic disconnection confidence of each triple data, calculating the fuzzy interference degree of each triple data, further obtaining the noise semantic feature value of each triple data to identify and eliminate noise triple data, and constructing a device operation and maintenance knowledge graph by using the retained triple data in the historical operation and maintenance data. The application can improve the construction quality of the device operation and maintenance knowledge graph.

Description

Technical Field

[0001] This application relates to the field of knowledge graph construction technology, specifically to an AI-based method and system for constructing a device operation and maintenance knowledge graph. Background Technology

[0002] Currently, building equipment operation and maintenance (O&M) knowledge graphs is a crucial method for improving O&M efficiency and quality. By utilizing AI technology, O&M knowledge graphs systematically display knowledge and information related to equipment O&M, including fault entities, entity relationships, and O&M events for various O&M objects. Furthermore, using O&M knowledge graphs helps teams quickly pinpoint the root cause of problems and find corresponding solutions, thereby achieving more efficient O&M management for various O&M objects. Therefore, O&M knowledge graphs, as an important tool in the field of equipment O&M, have significant value in improving the efficiency and quality of equipment O&M and reducing O&M costs.

[0003] In existing technologies, historical operation and maintenance data of various operation and maintenance objects are collected, and knowledge is extracted from this historical data using the DeepKE open-source knowledge extraction tool. Then, AI technology is used to construct an equipment operation and maintenance knowledge graph based on the extracted triples, aiming to improve the efficiency and quality of equipment operation and maintenance. However, noisy triples generated by fuzzy semantics in historical operation and maintenance data can cause anomalies in the semantic coherence between different triples. Existing technologies using the DeepKE open-source knowledge extraction tool cannot accurately uncover these anomalous features, resulting in poor accuracy in noise filtering of the extracted triples. This reduces the quality of the equipment operation and maintenance knowledge graph construction, impacting the efficiency and quality of equipment operation and maintenance. Summary of the Invention

[0004] To address the aforementioned technical issues, the purpose of this application is to provide an AI-based method and system for constructing a knowledge graph for equipment operation and maintenance. The specific technical solution adopted is as follows:

[0005] This application provides an AI-based method for constructing a device operation and maintenance knowledge graph, including the following steps:

[0006] The historical operation and maintenance data of each operation and maintenance object is obtained, and knowledge extraction is performed to obtain the data of each triple. Then, the text vector and text word vector of each triple data are extracted.

[0007] We analyze the similarity of text word vectors and the degree of semantic coherence among different triplet data to obtain the operational feature coherence of each triplet data. We also combine the local outlier features of the operational feature coherence of each triplet data to obtain the semantic disjointness confidence of each triplet data.

[0008] Based on the semantic similarity between different word segments in the text vector of each triplet data and the random variation of semantic similarity, the fuzzy interference degree of each triplet data is obtained, and then combined with the semantic disjoint confidence degree to obtain the noise semantic feature value of each triplet data.

[0009] Based on the semantic features of noise, noise triples are identified and removed, and then the triple data retained in the historical operation and maintenance data are used to construct a knowledge graph of equipment operation and maintenance.

[0010] Preferably, in the knowledge extraction process, the triplet data includes the subject, the object, and the relationship between the subject and the object.

[0011] Preferably, the similarity between the text word vectors of each triplet data and the text word vectors of each other triplet data is calculated, and the similarity between each triplet data and all other triplet data is clustered. The cluster corresponding to the cluster with the largest mean of all elements in the cluster is taken as the highly correlated cluster of each triplet data.

[0012] Preferably, the acquisition of the operational feature coherence of each triplet data further comprises:

[0013] ;

[0014] In the formula, The operational feature coherence of the i-th triplet data. For normalization function, Let be the number of elements within the highly correlated cluster of the i-th triplet. The sum of elements within the highly correlated cluster of the i-th triplet data. Let represent the degree of dispersion of elements within the highly correlated cluster of the i-th triplet data. To avoid constants with a denominator of 0.

[0015] Preferably, the reciprocal of the sum of the normalized results of the operational feature coherence degree and the local density of the operational feature coherence degree of each triplet data is used as the semantic disjointness confidence of each triplet data, wherein the local density of the operational feature coherence degree of each triplet data is obtained by the density peak clustering algorithm.

[0016] Preferably, the text vectors of each triplet data are segmented into words, and all the segmented words are combined into a segmented dataset for each triplet data.

[0017] Preferably, the fuzzy interference degree of each triplet data is obtained by:

[0018] ;

[0019] In the formula, Let represent the fuzzy interference degree of the i-th triplet data. It is an exponential function with the natural constant as its base. Let $\frac{i}{i}$ be the mean of the normalized Google distances among all words in the word segmentation dataset for the $i$-th triplet. Let be the information entropy of the normalized Google distance among all words in the word segmentation dataset for the i-th triplet data.

[0020] Preferably, the noise semantic feature value of each triplet data is the sum of the semantic disconnect confidence and the fuzzy interference degree of each triplet data.

[0021] This application also provides an AI-based device operation and maintenance knowledge graph construction system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described AI-based device operation and maintenance knowledge graph construction methods.

[0022] As can be seen from the above, the AI-based equipment operation and maintenance knowledge graph construction method and system provided in this application have at least the following beneficial effects:

[0023] This application extracts highly correlated clusters of triplet data through semantic similarity analysis, and accurately measures the coherence of operation and maintenance features between triplet data through data features within highly correlated clusters. This more clearly shows the correlation between equipment operation and maintenance information, retains triplet data information with good coherence of operation and maintenance features, and thus avoids erroneously filtering out important fault operation and maintenance information.

[0024] Furthermore, this application accurately reflects the degree of semantic coherence anomaly of the triplet data among all triplet data by extracting the local density features of the operational feature coherence of the triplet data. Then, it constructs a semantic disjoint confidence score by combining the operational feature coherence of the triplet data with its local density features, thereby clearly showing the problem of semantic disjointness or semantic incoherence between the triplet data and other triplet data, which can be used to more accurately filter out noisy triplets caused by ambiguous semantics in the future.

[0025] This application accurately measures the noisy semantic features of each triplet by assessing the credibility of semantic disconnect between the data and the degree of interference from ambiguous semantics. Then, it uses an anomaly detection algorithm to accurately identify and filter out noisy triplets, thereby improving the quality of equipment operation and maintenance knowledge graph construction and avoiding impacting the efficiency and quality of subsequent equipment operation and maintenance. Attached Figure Description

[0026] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 A flowchart illustrating the steps of an AI-based device operation and maintenance knowledge graph construction method provided in this application;

[0028] Figure 2 A system block diagram for constructing the equipment operation and maintenance knowledge graph provided in the embodiments of this application. Detailed Implementation

[0029] To further illustrate the technical means and effects adopted by this application to achieve the intended inventive purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an AI-based device operation and maintenance knowledge graph construction method and system proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0030] Unless otherwise specified and limited, terms such as “comprising,” “including,” or any other variations thereof are intended to cover a non-exclusive inclusion, such that a circuit structure, article, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the article or device that includes said element. Furthermore, the term “and / or” as used herein includes any and all combinations of one or more of the associated listed items. All technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0031] The following description, in conjunction with the accompanying drawings, details the specific scheme of the AI-based equipment operation and maintenance knowledge graph construction method and system provided in this application.

[0032] Please see Figure 1 The diagram illustrates a flowchart of an AI-based device operation and maintenance knowledge graph construction method according to an embodiment of this application, including the following steps:

[0033] Step 1: Obtain historical operation and maintenance data for each operation and maintenance object, perform knowledge extraction to obtain triple data, and then extract text vectors and text word vectors for each triple data.

[0034] To achieve more efficient equipment operation and maintenance management, it is necessary to accurately mine the abnormal features of semantic coherence between different triples in the knowledge extraction results, so as to accurately filter out noisy triples caused by fuzzy semantics, thereby improving the quality of equipment operation and maintenance knowledge graph construction.

[0035] Therefore, in this embodiment, historical operation and maintenance data of various operation and maintenance objects are collected first. The operation and maintenance objects include CPU servers, GPU servers, storage hard disks, and network devices. The historical operation and maintenance data is text data of failure cases and solutions in the historical process. Common operation and maintenance failures include CPU server overheating, GPU core failure (decreased computing power or GPU card failure), hard disk head damage, and network device port failure.

[0036] Knowledge extraction from historical operation and maintenance data is performed by installing and configuring the DeepKE open-source knowledge extraction tool on a computer. The knowledge extraction tasks include named entity recognition, relation extraction, attribute extraction, and triple data extraction, resulting in all triple data in the historical operation and maintenance data. The triple data contains the subject, the object, and the relationship between the subject and the object. The use of DeepKE for knowledge extraction is a well-known technology, and the specific process will not be described in detail.

[0037] Furthermore, the knowledge extraction triplet data is preprocessed based on AI artificial intelligence technology. Specifically, each triplet data in the historical operation and maintenance data is transformed into a text vector to obtain a text vector for each triplet data. The text vector of each triplet data is then used as input to the Word2Vec word vector algorithm to obtain the text word vector of each triplet data. The text vector transformation and the Word2Vec word vector algorithm are well-known technologies in the field of AI artificial intelligence, and the specific process will not be elaborated here.

[0038] Step 2: Analyze the similarity of text word vectors and the degree of semantic coherence among different triplet data to obtain the operational feature coherence of each triplet data. Combine the local outlier features of the operational feature coherence of each triplet data to obtain the semantic disjointness confidence of each triplet data.

[0039] Noisy triples arising from ambiguous semantics exist in historical operation and maintenance data. Current technologies, such as the DeepKE open-source knowledge extraction tool, cannot accurately uncover anomalous features related to semantic coherence between different triples, thus failing to effectively filter out these noisy triples. Therefore, to improve the quality of equipment operation and maintenance knowledge graph construction, it is necessary to fully explore anomalous features related to semantic coherence between different triples.

[0040] To analyze the correlation characteristics between different triplet data in historical operation and maintenance data, taking the i-th triplet data as an example, the similarity between the text word vector of the i-th triplet data and the text word vector of each of the other triplet data is calculated. The similarity measurement method can be cosine similarity or Jaccard similarity coefficient. In this embodiment, cosine similarity is used to measure the similarity. The larger the similarity, the higher the similarity between the text word vectors. The measurement of cosine similarity is a well-known technique, and the specific process will not be described in detail.

[0041] Furthermore, to further analyze the semantic coherence features between each triplet and other triplets, the similarity between the i-th triplet and all other triplets is used as the input to the K-means clustering algorithm. The elbow rule is used to obtain the optimal number of clusters K, and the K-means clustering algorithm is used to obtain K clusters composed of similarity. The mean of all elements in each cluster is calculated, and the cluster corresponding to the maximum mean is taken as the highly correlated cluster of the i-th triplet. The K-means clustering algorithm and the elbow rule are well-known techniques, and the specific process will not be described in detail.

[0042] Generally, the more elements in a highly correlated cluster of a triplet, the more information related to that triplet is in the historical operation and maintenance data. Therefore, the triplet is less likely to be a noisy triplet generated by fuzzy semantics.

[0043] Furthermore, the larger the sum of elements within a highly correlated cluster and the smaller the dispersion, the better the consistency of operational characteristics between some triplet data in historical operational data and the actual triplet data. Therefore, the triplet data information is more important for constructing the operational knowledge graph. For example, if the triplet data contains faulty entities in equipment operation and maintenance, such as damaged hard drive heads or packet loss at switch ports, then the corresponding solutions contain numerous triplet data pieces closely related to these faulty entities.

[0044] Based on the above analysis, the operational characteristic coherence degree of each triplet data is calculated:

[0045] ;

[0046] In the formula, The operational feature coherence of the i-th triplet data. For normalization function, Let be the number of elements within the highly correlated cluster of the i-th triplet. The sum of elements within the highly correlated cluster of the i-th triplet data. Let represent the degree of dispersion of elements within the highly correlated cluster of the i-th triplet data. To avoid constants with a denominator of 0, the value is taken within a small data range (0.01, 0.1), and the impact on the calculation result can be ignored. In this embodiment, the adjustment parameter is set to 0.05.

[0047] The method for measuring the degree of dispersion can be variance, standard deviation, or coefficient of dispersion. In this embodiment, the coefficient of dispersion is used to measure the degree of dispersion.

[0048] Based on the above calculation process, it can be understood that the consistency of operation and maintenance features reflects the consistency of operation and maintenance features between the i-th triplet data and other triplet data. The greater the consistency of operation and maintenance features, the better the consistency of operation and maintenance features between the triplet data and other triplet data in historical operation and maintenance data. In this case, the triplet data is less likely to be a noisy triplet generated by fuzzy semantics, and should be retained when constructing the equipment operation and maintenance knowledge graph.

[0049] Furthermore, to accurately extract anomalous features of semantic coherence among different triplet data, the coherence of operational features of all triplet data in historical operational data is used as the basis for the DPC density peak clustering algorithm. The input to the peak clustering (DPC) algorithm is a data point clustering algorithm. The cutoff distance is selected according to the rule that the number of data points within the cutoff distance of each data point accounts for 1%-2% of the total number of data points. In this embodiment, the cutoff distance when the number of data points within the cutoff distance of each data point accounts for 2% of the total number of data points is used as the preset cutoff distance in the algorithm. The local density of the operational feature coherence of each triplet data is obtained by the DPC density peak clustering algorithm. The smaller the local density of the operational feature coherence, the more significant the outlier feature of the corresponding triplet data in the historical operational data. That is, the higher the degree of semantic coherence abnormality of the triplet data in all triplet data, the more likely the triplet data belongs to the noisy triplet generated by fuzzy semantics. Therefore, the triplet data should be filtered out when constructing the equipment operation and maintenance knowledge graph to avoid affecting the quality of the equipment operation and maintenance knowledge graph construction.

[0050] Therefore, in order to more accurately filter out noisy triples caused by ambiguous semantics, the local density of the operational feature coherence of all triples in the historical operational data is normalized to the maximum value. The reciprocal of the sum of the operational feature coherence of each triple and the normalized result of the local density of the operational feature coherence is used as the semantic disjointness confidence of each triple.

[0051] Understandably, the semantic disjointness confidence level reflects the degree of confidence in the loss of semantic coherence between the triplet data and other triplet data in historical operation and maintenance data. The higher the confidence level of the loss of semantic coherence, the more likely it is that the semantic relationship between the triplet data and other triplets has a semantic disjointness or semantic incoherence. In this case, the triplet data is more likely to be a noisy triplet generated by fuzzy semantics, which will affect the quality of the equipment operation and maintenance knowledge graph construction. Therefore, it is necessary to filter out noisy triplets generated by fuzzy semantics.

[0052] Step 3: Based on the semantic similarity between different word segments in the text vector of each triplet data and the random variation of the semantic similarity, obtain the fuzzy interference degree of each triplet data, and then combine it with the semantic disjoint confidence to obtain the noise semantic feature value of each triplet data.

[0053] Furthermore, in order to extract the features of fuzzy semantics within the text vectors of noisy triples, the text vector of each triple data is used as the input of Jieba segmentation technology. Jieba segmentation technology is used to segment the text vector of each triple data, and the set of all segmented words after segmentation is used as the segmentation dataset of each triple data. Jieba segmentation technology is a well-known technology, and the specific process will not be described in detail.

[0054] Generally, the semantic similarity between different words in a normal triplet data set is relatively high. However, in noisy triplet data generated by noise or fuzzy semantics, the semantic similarity between different words in the noisy triplet data set is low due to the interference of fuzzy semantics, and the semantic similarity between different words is relatively chaotic. Therefore, noisy triplet data should be filtered out when constructing a device operation and maintenance knowledge graph to avoid affecting the quality of the knowledge graph construction. Thus, in this embodiment, the fuzzy interference degree of each triplet data is calculated based on the average level of Google distance between different words in the triplet data and the random variation of Google distance:

[0055] ;

[0056] In the formula, Let represent the fuzzy interference degree of the i-th triplet data. It is an exponential function with the natural constant as its base. Let $\frac{i}{i}$ be the mean of the normalized Google distances among all words in the word segmentation dataset for the $i$-th triplet. Let be the information entropy of the normalized Google distance among all words in the word segmentation dataset for the i-th triplet data.

[0057] Based on the above calculation process, it can be understood that the fuzzy interference degree reflects the degree of interference of each triplet data with fuzzy semantics. The greater the fuzzy interference degree, the more likely the subject, object, and the relationship between subject and object in the corresponding triplet data are to be affected by fuzzy semantics. Therefore, the corresponding triplet data is more likely to be a noisy triplet generated by fuzzy semantics.

[0058] Furthermore, by combining the semantic disjointness features and fuzzy interference features of the triplet data, the noise semantic features of each triplet data are measured, thereby enabling accurate filtering of noisy triplets generated by fuzzy semantics. Specifically, the sum of the semantic disjointness confidence and fuzzy interference degree of each triplet data in historical operation and maintenance data is used as the noise semantic feature value of each triplet data. The noise semantic feature value reflects the degree to which each triplet data is affected by noise. The higher the anomaly of a triplet data's noise semantic feature value among all triplet data's noise semantic feature values, the more severe the noise impact on that triplet data, and the more likely it is to be filtered out, thus avoiding affecting the quality of the equipment operation and maintenance knowledge graph construction.

[0059] Step 4: Identify and remove noise triples based on noise semantic feature values, and then use the triple data retained in historical operation and maintenance data to construct an equipment operation and maintenance knowledge graph.

[0060] Furthermore, in order to accurately filter out noisy triples generated by ambiguous semantics, the noisy semantic feature values ​​of all triple data in the historical operation and maintenance data are used as the input of the LOF anomaly detection algorithm (Local Outlier Factor). The preset neighborhood parameter in the algorithm is 10. The LOF anomaly detection algorithm detects abnormal data within all noisy semantic feature values. The LOF anomaly detection algorithm is a well-known technology, and the specific process will not be described in detail.

[0061] In this embodiment, for all triplet data, the above process can be used to perform anomaly detection on the noise semantic feature values ​​of all triplet data, extract the abnormal data in the noise semantic feature values, and record the triplet data corresponding to the abnormal data in all noise semantic feature values ​​as noise triplet. It should be noted that the noise triplet is the triplet data corresponding to the abnormal data in the anomaly detection result of the noise semantic feature values.

[0062] By removing noisy triples from all triple data, noisy triples generated by fuzzy semantics are filtered out, resulting in filtered historical operation and maintenance data. Then, knowledge fusion and knowledge graph construction are performed on the filtered historical operation and maintenance data using the Protege open-source software to obtain the equipment operation and maintenance knowledge graph. The use of Protege open-source software for knowledge fusion and knowledge graph construction is a well-known technology, and the specific process will not be elaborated here.

[0063] Furthermore, in this embodiment, the equipment operation and maintenance knowledge graph is stored in the Neo4j graph database in the form of graph data. The Neo4j graph database serves as a graph database engine and stores knowledge in the form of graph data. Fault entities in equipment operation and maintenance are represented as nodes, and relationships exist in the form of edges. This graph database storage method provides efficient equipment operation and maintenance fault query and fault correlation analysis capabilities, enabling equipment operation and maintenance personnel to easily perform complex equipment operation and maintenance operations, thereby better understanding equipment operation and maintenance faults and fault resolution methods.

[0064] Based on the same inventive concept as the above method, this application embodiment also provides an AI-based device operation and maintenance knowledge graph construction system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described AI-based device operation and maintenance knowledge graph construction methods.

[0065] Preferably, in this embodiment, the equipment operation and maintenance knowledge graph construction system includes a data processing module, a coherence analysis module, a noise analysis module, and a knowledge graph construction module. The data processing module acquires historical operation and maintenance data for each operation and maintenance object, performs knowledge extraction to obtain triplet data, and then extracts text vectors and text word vectors from each triplet data. The coherence analysis module analyzes the similarity of text word vectors and the degree of semantic coherence between different triplet data to obtain the operation and maintenance feature coherence of each triplet data. It also combines the local outlier features of the operation and maintenance feature coherence of each triplet data to obtain the semantic disjointness confidence level of each triplet data. The noise analysis module obtains the fuzzy interference degree of each triplet data based on the semantic similarity between different word segments in the text vectors of each triplet data and the random variation of the semantic similarity. It then combines the semantic disjointness confidence level to obtain the noise semantic feature value of each triplet data. The knowledge graph construction module identifies and removes noisy triples based on the noise semantic feature value, and then constructs an equipment operation and maintenance knowledge graph using the triplet data retained from the historical operation and maintenance data. Specifically, the system block diagram for constructing the equipment operation and maintenance knowledge graph provided in this embodiment is as follows: Figure 2 As shown.

[0066] It is understood that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0067] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0068] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the protection scope of this application.

Claims

1. A method for constructing an AI-based equipment operation and maintenance knowledge graph, characterized in that, Includes the following steps: The historical operation and maintenance data of each operation and maintenance object is obtained, and knowledge extraction is performed to obtain the data of each triple. Then, the text vector and text word vector of each triple data are extracted. We analyze the similarity of text word vectors and the degree of semantic coherence among different triplet data to obtain the operational feature coherence of each triplet data. We also combine the local outlier features of the operational feature coherence of each triplet data to obtain the semantic disjointness confidence of each triplet data. Based on the semantic similarity between different word segments in the text vector of each triplet data and the random variation of semantic similarity, the fuzzy interference degree of each triplet data is obtained, and then combined with the semantic disjoint confidence degree to obtain the noise semantic feature value of each triplet data. Based on the semantic features of noise, noise triples are identified and removed, and then the triple data retained in the historical operation and maintenance data are used to construct an equipment operation and maintenance knowledge graph. The similarity between the text word vectors of each triplet data and the text word vectors of every other triplet data is calculated. Then, the similarity between each triplet data and all other triplet data is clustered. The cluster with the largest mean of all elements within the cluster is taken as the highly correlated cluster of each triplet data. The acquisition of the operational feature coherence of each triplet data is further as follows: In the formula, The operational feature coherence of the i-th triplet data. For normalization function, Let be the number of elements within the highly correlated cluster of the i-th triplet. The sum of elements within the highly correlated cluster of the i-th triplet data. Let represent the degree of dispersion of elements within the highly correlated cluster of the i-th triplet data. To avoid constants with a denominator of 0.

2. The method for constructing an AI-based equipment operation and maintenance knowledge graph as described in claim 1, characterized in that, In knowledge extraction, triplet data contains the subject, the object, and the relationship between the subject and the object.

3. The method for constructing an AI-based equipment operation and maintenance knowledge graph as described in claim 1, characterized in that, The reciprocal of the sum of the normalized results of the operational feature coherence of each triplet data and the local density of the operational feature coherence is used as the semantic disjointness confidence of each triplet data. The local density of the operational feature coherence of each triplet data is obtained by the density peak clustering algorithm.

4. The method for constructing an AI-based equipment operation and maintenance knowledge graph as described in claim 1, characterized in that, The text vectors of each triplet data are segmented into words, and all the segmented words are combined into a segmented dataset for each triplet data.

5. The method for constructing an AI-based equipment operation and maintenance knowledge graph as described in claim 4, characterized in that, The fuzzy interference degree of each triplet data is further obtained as follows: In the formula, Let represent the fuzzy interference degree of the i-th triplet data. It is an exponential function with the natural constant as its base. Let $\frac{i}{i}$ be the mean of the normalized Google distances among all words in the word segmentation dataset for the $i$-th triplet. Let be the information entropy of the normalized Google distance among all words in the word segmentation dataset for the i-th triplet data.

6. The method for constructing an AI-based equipment operation and maintenance knowledge graph as described in claim 1, characterized in that, The noise semantic feature value of each triplet data is the sum of the semantic disconnect confidence and fuzzy interference of each triplet data.

7. The method for constructing an AI-based equipment operation and maintenance knowledge graph as described in claim 1, characterized in that, Anomaly detection is performed on the noise semantic feature values ​​of all triplet data, and the triplet data corresponding to the abnormal data is taken as the noise triplet.

8. An AI-based device operation and maintenance knowledge graph construction system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the AI-based device operation and maintenance knowledge graph construction method as described in any one of claims 1-7.

Images