Equipment fault diagnosis recommendation method based on knowledge graph and multi-dimensional relevance indexes

Abstract

The invention provides an equipment fault diagnosis recommendation method based on a knowledge graph and a multi-dimensional relevance index, and the method comprises the steps: converting a fault text into triple structured data based on an equipment fault knowledge graph ontology model, and constructing an equipment fault knowledge graph; constructing a multi-dimensional correlation index of the fault mode, and converting explicit feedback in the historical fault diagnosis data into implicit feedback; obtaining a fault mode which is strongly correlated with the abnormal phenomenon from historical fault diagnosis data based on a multi-dimensional correlation index, and taking the fault mode as a seed fault mode; and taking the seed fault mode as a starting point, iteratively propagating in the fault knowledge graph to obtain a ripple set, calculating an association probability between an abnormal phenomenon and a candidate fault mode according to the ripple set, accumulating each order of response according to the association probability, and outputting a prediction definite diagnosis probability. According to the method, the diagnosis relation strength between the abnormal phenomenon and the fault mode is revealed according to the multi-dimensional relevance index, and a flexible reasoning mechanism is adopted on the basis, so that the diagnosis accuracy and adaptability are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of equipment fault diagnosis, and relates to an equipment fault diagnosis recommendation method based on a knowledge graph and multi-dimensional correlation indexes. Background Art

[0002] Equipment refers to mechanical equipment and systems with specific functions for performing special tasks, which are widely used in key fields such as national defense, manufacturing, and transportation, and undertake the functions of ensuring tasks, improving production efficiency, and performing specialized operations. During the operation of equipment, effective diagnosis and timely disposal of faults are directly related to its operation efficiency and the reliability of task completion.

[0003] In recent years, knowledge graphs have received extensive attention due to their information integration and reasoning capabilities in fault diagnosis, and many researchers have conducted in-depth explorations in this direction. However, there are still some deficiencies in the current diagnosis recommendation methods based on knowledge graphs when applied to complex equipment fault diagnosis:

[0004] First, the fault knowledge graphs constructed in existing research are often relatively simple in structure, mainly focusing on machine factors, and insufficiently considering the two key factors of people and environment, making it difficult to comprehensively reveal the root cause relationship of complex faults.

[0005] Second, traditional knowledge graph embedding and path-based reasoning methods have insufficient interpretability of recommendation results in practical applications, and are severely limited by predefined meta-paths, making it difficult to meet the requirements of equipment diagnosis for flexibility and comprehensiveness.

[0006] Third, to solve the problem of insufficient interpretability of recommendation results, the RippleNet recommendation algorithm based on a mixture of knowledge graph embedding and paths can be considered. However, when the existing RippleNet recommendation algorithm is applied to fault diagnosis, it is difficult to accurately reflect the correlation between different fault modes and abnormal phenomena, further affecting the diagnosis accuracy.

[0007] Therefore, how to provide an equipment fault diagnosis recommendation method that can further improve the diagnosis accuracy is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0008] In view of this, the present invention proposes an equipment fault diagnosis recommendation method based on a knowledge graph and multi-dimensional correlation indicators, which reveals the strength of the diagnostic relationship between abnormal phenomena and fault modes in historical diagnosis information. On this basis, a flexible reasoning mechanism is adopted to iteratively propagate the fault information in the knowledge graph, layer by layer expanding and dynamically mining the complex relationship between the emerging abnormal phenomena and potential fault modes. This iterative propagation method enables the fault modes to be deeply mined at a wider level of the graph, thereby improving the accuracy and adaptability of diagnosis and meeting the actual needs of complex equipment fault diagnosis.

[0009] To achieve the above object, the present invention adopts the following technical solutions:

[0010] The present invention discloses an equipment fault diagnosis recommendation method based on a knowledge graph and multi-dimensional correlation indicators, comprising the following steps:

[0011] S1: Based on the abnormal phenomena of equipment units and the fault logics of human factors, machine factors, and environmental factors, define the semantic types and semantic relationships in the equipment fault domain, and construct an equipment fault knowledge graph ontology model;

[0012] S2: Obtain fault texts from equipment fault cases, and based on the semantic types and semantic relationships in the equipment fault knowledge graph ontology model, use a knowledge extraction algorithm to convert the fault texts into triple-structured data, and construct an equipment fault knowledge graph; the equipment fault cases include historical fault diagnosis data;

[0013] S3: Construct multi-dimensional correlation indicators corresponding to abnormal phenomena and fault modes, use the multi-dimensional correlation indicators to optimize the correlation representation between abnormal phenomena and fault modes, and convert the explicit feedback in historical fault diagnosis data into implicit feedback;

[0014] S4: Receive the input abnormal phenomena, and based on the correlation representation, obtain the fault modes whose correlation satisfies the preset conditions from the historical fault diagnosis data as seed fault modes; construct a diagnosis recommendation model, starting from the seed fault modes, perform iterative propagation in the fault knowledge graph to obtain a ripple set, obtain the fault modes in the propagation process according to the ripple set, and use the seed fault modes and the fault modes in the propagation process as candidate fault modes; calculate the correlation probability between the input abnormal phenomena and the candidate fault modes according to the ripple set, accumulate the responses of each order according to the correlation probability, output the predicted diagnosis probability, and determine the diagnosis result.

[0015] Preferably, the equipment fault knowledge graph ontology model has a multi-level structure, including:

[0016] The unit layer is used to describe the spatial positions and structural association information of equipment, systems, and components;

[0017] A logic layer, which is used to describe fault manifestations and their mechanisms, and analyze fault causes, including any one or a combination of more than one of the following: human factors, machine factors, and environmental factors;

[0018] A maintenance layer, which is used to define maintenance measures and solutions.

[0019] Preferably, the semantic types in the ontology model of the equipment fault knowledge graph include any combination of more than one of the following: equipment name, system name, component name, fault mode, abnormal phenomenon, human factor, machine factor, or environmental factor; the semantic relationships include any combination of more than one of the following: composition, attribution, location, cause, inducement, manifestation, or adoption.

[0020] Preferably, the equipment fault cases in S2 include structured text information and unstructured text information; among them,

[0021] The structured text information includes historical fault data;

[0022] The unstructured text information includes professional technical documents, fault troubleshooting records, and maintenance manuals.

[0023] Preferably, a pre-trained BERT-BiLSTM-CRF model is used to perform named entity recognition on the unstructured text information; a pre-trained Chinese-BERT-wwm-ext model is used to extract relationships between fault-related entities from the unstructured text information.

[0024] Preferably, S3 includes: obtaining abnormal phenomena and corresponding diagnosed fault modes, the fault frequency, fault level, and influence range corresponding to the fault mode from historical fault diagnosis data as multi-dimensional correlation indicators; among them,

[0025] The fault frequency is used to record the number of times the abnormal phenomenon is diagnosed as the corresponding fault mode;

[0026] The fault level is used to evaluate the severity of the fault according to the influence degree of the fault mode;

[0027] The influence range is used to measure the impact of the fault mode on each component in the system.

[0028] Preferably, the steps of using the multi-dimensional correlation indicators in S3 to optimize the correlation representation between abnormal phenomena and fault modes include:

[0029] Standardizing the multi-dimensional correlation indicators to the same dimension;

[0030] Substitute the standardized multi-dimensional correlation indicators into the calculation formula of the Fault Mode Association Evaluation Index FMAEI, which is used to measure the association strength between abnormal phenomena and fault modes. The calculation formula is as follows:

[0031] ;

[0032] In the formula, α1, α2, and α3 are the weight coefficients of the three dimensions of fault frequency, fault level, and influence range respectively, and the sum of the weight coefficients is 1; S f , Ss, S i are the fault frequency, fault level, and influence range data after standardization respectively.

[0033] Preferably, a diagnostic recommendation model is constructed based on the RippleNet recommendation algorithm in S4.

[0034] Preferably, S4 includes the following steps:

[0035] S41: Calculate the association representation between the abnormal phenomenon and the fault mode according to the historical fault diagnosis data, and obtain the fault mode that satisfies the preset conditions historically for the input abnormal phenomenon as the seed fault mode;

[0036] S42: Put the seed fault mode into the equipment fault knowledge graph as the starting point, obtain the relevant ripple set through iterative propagation, obtain the fault mode in the propagation process according to the ripple set, and use the seed fault mode and the fault mode in the propagation process as candidate fault modes;

[0037] S43: Calculate the association probability between the input abnormal phenomenon and the candidate fault mode according to the head entity and relationship of the ripple set;

[0038] S44: Accumulate the responses of each order of the ripple set layer by layer according to the association probability, accumulate the responses of each order, combine the corresponding candidate fault modes, output the predicted diagnosis probability, and determine the diagnosis result according to the predicted diagnosis probability.

[0039] Preferably, the calculation steps of the predicted diagnosis probability between the abnormal phenomenon and the candidate fault mode in S4 include:

[0040] Obtain the 1-hop ripple set of the candidate fault mode v and the input abnormal phenomenon w in the equipment fault knowledge graph , by comparing the candidate fault mode v with the head entity and the relationship in the corresponding triple, assign an association probability to each triple in :

[0041] ;

[0042] In the formula, , is the embedding of relationship r i and head entity h i . is the embedding of candidate fault mode v, d is the dimension of the embedding, p i represents the correlation degree between v and h i , and i is the number of triples in; softmax() is the softmax function;

[0043] Take all the tail entities in and sum them up with corresponding association probability p i to obtain the diagnostic history V of abnormal phenomenon w w For the first-order response vector of candidate fault mode v :

[0044] ;

[0045] In the formula, is the embedding of tail entity t i .

[0046] For the first-order response vector perform iterative propagation in the ripple set of abnormal phenomenon w to obtain the high-order response of abnormal phenomenon w , and the response of abnormal phenomenon w to candidate fault mode v is expressed as the sum of each-order response:

[0047] ;

[0048] In the formula, is the H-th order response of abnormal phenomenon w, and H is a positive integer greater than 2;

[0049] Combining abnormal phenomenon w and candidate fault mode v, obtain the predicted probability:

[0050] ;

[0051] In the formula, is the sigmoid function, is the embedding of abnormal phenomenon w;

[0052] According to the predicted probability, screen all candidate fault modes to obtain a recommended list of diagnosis results.

[0053] It can be seen from the above technical solutions that compared with the existing method of realizing diagnosis by simple search and matching depending on predefined inference paths, the beneficial effects of the present invention include:

[0054] (1) In the process of constructing the equipment fault knowledge graph of the present invention, by introducing multi-dimensional information of the three elements of human, machine, and environment, the comprehensiveness and accuracy of the graph are effectively enhanced, making the fault diagnosis more in line with the actual application scenario, and making up for the deficiency that the traditional knowledge graph focuses too much on machine factors.

[0055] (2) The present invention provides a more detailed and comprehensive evaluation of the relationship between fault modes and abnormal phenomena through multi-dimensional correlation indicators, dynamically revealing the strength of the correlation between abnormal phenomena and fault modes in historical diagnosis information, and avoiding the limitation of insufficient correlation recognition in traditional methods.

[0056] (3) Through the diagnosis recommendation model of the present invention, the input abnormal phenomena are automatically associated with strongly related fault modes, and in-depth analysis is carried out relying on the multi-level association relationships in the knowledge graph. This not only effectively reduces the burden of maintenance personnel for fault troubleshooting, but also improves the diagnosis speed and accuracy through a flexible reasoning mechanism, greatly enhancing the efficiency and applicability of equipment fault diagnosis. Brief Description of the Drawings

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings;

[0058] Figure 1 It is a flowchart of the equipment fault diagnosis recommendation method based on the knowledge graph and multi-dimensional correlation indicators provided by the embodiment of the present invention;

[0059] Figure 2 It is a schematic structural diagram of the equipment fault knowledge graph ontology model provided by the embodiment of the present invention;

[0060] Figure 3 It is a schematic diagram of historical diagnosis information for the abnormal phenomenon of "insufficient power output" provided by the embodiment of the present invention;

[0061] Figure 4 It is a schematic diagram of the iterative propagation of the "fault of injector 1" fault mode in the fault knowledge graph provided by the embodiment of the present invention. Detailed Embodiment

[0062] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0063] As Figure 1 shown, the embodiment of the present invention provides an equipment fault diagnosis recommendation method based on a knowledge graph and multi-dimensional correlation indicators. First, a multi-level equipment fault knowledge graph that comprehensively considers the three elements of people, machines, and the environment is constructed. Secondly, a fault mode correlation evaluation index containing multi-dimensional fault characteristics is used to transform the explicit feedback in historical diagnosis information. Finally, a diagnosis recommendation model is constructed by combining the RippleNet recommendation algorithm with high efficient association reasoning ability, so as to realize the accurate diagnosis of equipment faults.

[0064] It includes the following steps:

[0065] S1: Based on the abnormal phenomena of equipment units and the fault logics of human factors, machine factors, and environmental factors, define the semantic types and semantic relationships in the equipment fault domain, and construct an ontology model of the equipment fault knowledge graph;

[0066] S2: Obtain fault texts from equipment fault cases, and based on the semantic types and semantic relationships in the ontology model of the equipment fault knowledge graph, use a knowledge extraction algorithm to transform the fault texts into triple structured data, and construct an equipment fault knowledge graph; the equipment fault cases include historical fault diagnosis data;

[0067] S3: Construct multi-dimensional correlation indicators for the fault modes corresponding to abnormal phenomena, use the multi-dimensional correlation indicators to optimize the correlation representation between abnormal phenomena and fault modes, and transform the explicit feedback in historical fault diagnosis data into implicit feedback;

[0068] S4: Receive the input abnormal phenomenon, and based on the correlation representation, obtain the fault modes whose correlations meet the preset conditions from the historical fault diagnosis data as seed fault modes; construct a diagnosis recommendation model, start from the seed fault modes, perform iterative propagation in the fault knowledge graph to obtain a ripple set, obtain the fault modes in the propagation process according to the ripple set, and use the seed fault modes and the fault modes in the propagation process as candidate fault modes; calculate the correlation probability between the input abnormal phenomenon and the candidate fault modes according to the ripple set, accumulate the responses of each order according to the correlation probability, output the predicted diagnosis probability, and determine the diagnosis result.

[0069] In one embodiment, equipment fault diagnosis involves a large amount of complex and diverse knowledge, such as the composition structure of the equipment, fault modes and their causes, etc. Organizing this knowledge in a computable and inferable manner requires relying on precise ontology construction. To systematically and comprehensively describe equipment faults and their related information, starting from three key issues for the occurred fault events: "Where is this fault located?", "Why does this fault occur?", "What maintenance measures should be taken for this fault?", an ontology structure including three levels of unit, logic, and maintenance is constructed. Among them:

[0070] The unit layer is used to describe the spatial location and structural association information of equipment, systems, and components;

[0071] The logic layer is used to describe fault manifestations and their mechanisms, and analyze fault causes, including any one or a combination of the following: human factors, machine factors, and environmental factors;

[0072] The maintenance layer is used to define maintenance measures and solutions.

[0073] In this embodiment, considering that the occurrence of equipment faults is usually the result of the intertwined action of multiple reasons, therefore, in the analysis of the root causes of faults, from the perspective of the "human-machine-environment" systems engineering, the three major elements of human factors, machine factors, and environmental factors are unified in the equipment fault knowledge graph for modeling, comprehensively revealing the causes of faults, and providing a scientific basis and means for equipment fault diagnosis and maintenance. The overall equipment fault knowledge level division is shown in Table 1, and the relationships between each semantic type are as Figure 2 shown.

[0074] Table 1 Equipment Fault Knowledge Level Division

[0075]

[0076] In one embodiment, the semantic types in the ontology model of the equipment fault knowledge graph include any combination of the following: equipment name, system name, component name, fault mode, abnormal phenomenon, human factor, machine factor, or environmental factor; the semantic relationships include any combination of the following: composition, attribution, location, cause, inducement, manifestation, or take.

[0077] In one embodiment, the equipment fault cases in S2 include structured text information and unstructured text information; among them,

[0078] The structured text information includes historical fault data. The historical fault data is stored in the form of a table, containing information such as fault modes, faulty components, fault causes, and corresponding maintenance measures. Since these data have clear structured features, they can be directly converted into triples in the knowledge graph.

[0079] It should be noted that the historical fault diagnosis data in equipment failure cases also uses structured data, but it does not participate in the construction of the fault knowledge graph. Its role is to serve as the judgment basis for diagnosis recommendations together with the knowledge graph recommendation data.

[0080] Unstructured text information includes professional technical documents, troubleshooting records, and maintenance manuals. Professional technical documents cover equipment instructions, technical standards, and other content, and describe the structure and working principle of the equipment in detail; troubleshooting records are logs generated by on-site technicians during the diagnosis and troubleshooting process; maintenance manuals provide operating specifications and solutions for equipment maintenance and repair. Professional technical documents, troubleshooting records, maintenance manuals, and other data contain a large amount of unstructured text information, which needs to be converted into structured data through knowledge extraction technology to provide a basis for the subsequent construction of knowledge graphs.

[0081] It is necessary to explain the historical fault diagnosis data and historical fault data:

[0082] Historical fault data refers to a structured knowledge base centered on fault modes, including abnormal phenomena, fault modes, faulty components, the system to which they belong, fault levels, fault causes, and corresponding maintenance measures, as shown in Table 2.

[0083] Table 2 Example of historical fault dataset

[0084]

[0085] Historical fault diagnosis information data refers to the information obtained by diagnosing a certain abnormal phenomenon, including abnormal phenomenon, fault mode, fault frequency, fault level, and impact range, which are stored in a five-column table. Among them, "fault frequency, fault level, and impact range" are explicit feedback, as shown in Table 3.

[0086] Table 3. Example of historical fault diagnosis dataset

[0087]

[0088] The training data of the diagnosis recommendation model includes: knowledge graph triple structure data and historical fault diagnosis data.

[0089] In one embodiment, for unstructured fault texts such as troubleshooting records and maintenance manuals, named entity recognition and relationship extraction are performed through a deep learning model to obtain structured equipment fault-related triple data. In this embodiment, a pre-trained BERT-BiLSTM-CRF model is used to perform named entity recognition on unstructured text information; and a pre-trained Chinese-BERT-wwm-ext model is used to extract relationships between fault-related entities in unstructured text information.

[0090] In this embodiment, the BERT-BiLSTM-CRF model is used for named entity recognition, and the specific steps are as follows:

[0091] S201: Use an annotation tool to annotate the fault text, and adopt the BIO annotation method to annotate the entity boundaries in the text. The BIO method can effectively mark whether each word is part of an entity, and clarify the start position (B), middle position (I), or non-entity position (O) of the entity, forming a standardized annotation dataset.

[0092] S202: Input the annotation dataset into the BERT-BiLSTM-CRF model for training. The BERT layer generates the context semantic representation of words, the BiLSTM layer further extracts sequence features, and the CRF layer optimizes the entity boundaries to ensure the consistency and accuracy of the entity recognition results.

[0093] S203: Adopt accuracy, recall, and F1 value as evaluation indicators, and improve the robustness and generalization ability of the model by adjusting the hyperparameters of the model. Among them, the F1 value is an index used to measure the precision of a binary classification model in information retrieval and statistics, and it is defined as the harmonic mean of accuracy (Precision) and recall (Recall).

[0094] S204: Apply the trained model to new fault texts, automatically extract entity data from them, and organize them into a structured data table to provide data support for knowledge graph construction.

[0095] In this embodiment, the Chinese-BERT-wwm-ext model is used for relationship extraction between fault-related entities, and the specific steps are as follows:

[0096] S211: Perform relationship annotation on the entity pairs in the training data to form a relationship annotation dataset and provide training data for model learning.

[0097] S212: Use the Chinese-BERT-wwm-ext model for training, enhance the model's understanding ability of semantic associations in the text through the whole-word masking technique, and perform relationship classification learning on the entity pairs in the annotation data.

[0098] S213: Apply the trained model to new fault texts, automatically identify the association relationships between entities from them, and output structured triples containing "entity 1 - relationship - entity 2" to provide high-quality data support for knowledge graph construction.

[0099] In one embodiment, to efficiently manage the constructed equipment fault knowledge graph, this embodiment uses the Neo4j graph database as the storage system. Specifically, the structured fault data obtained through knowledge extraction will be saved in the.CSV file format, and then written into the Neo4j graph database using the Cypher language to construct the equipment fault knowledge graph through Neo4j.

[0100] The Neo4j graph database stores data using a graph structure (nodes and edges), which can represent entities and their relationships more intuitively and flexibly compared to traditional relational databases. It is particularly suitable for representing complex multi-dimensional relationships and can significantly improve the efficiency of querying and reasoning.

[0101] Through the Cypher language, the structured fault triple data is written into the Neo4j graph database to form the equipment fault knowledge graph. The Cypher language has efficient and intuitive graph data operation capabilities, can support complex query statements, and is convenient for in-depth querying of fault modes, fault causes, and their associated information. Utilizing the characteristics of the graph database, the querying of fault-related information after obtaining the diagnostic results can be more efficient, thereby ensuring that the root cause and associated information of equipment faults can be quickly found.

[0102] In one embodiment, S3 includes: obtaining abnormal phenomena and corresponding diagnosed fault modes from historical fault diagnosis data, and using the fault frequency, fault level, and influence scope corresponding to the fault mode as multi-dimensional correlation indicators; among them,

[0103] The fault frequency is used to record the number of times an abnormal phenomenon is diagnosed as the corresponding fault mode;

[0104] The fault level is used to evaluate the severity of a fault based on the degree of influence of the fault mode (such as the downtime and repair difficulty caused by the fault);

[0105] The influence scope is used to measure the impact of a fault mode on each component in the system.

[0106] In one embodiment, the purpose of S3 is to calculate the Fault Mode Association Evaluation Index (FMAEI). Through this index, the association strength between abnormal phenomena and fault modes is revealed, so as to provide accurate prior knowledge for the diagnostic recommendation model. The steps of optimizing the association representation between abnormal phenomena and fault modes using multi-dimensional correlation indicators include:

[0107] Standardization processing step:

[0108] After obtaining the historical fault diagnosis data, standardization processing will be performed to ensure that data in different dimensions have a unified scale and comparability. First, calculate the mean μ and standard deviation σ of each dimension, and the calculation formulas are as follows:

[0109] ;

[0110] ;

[0111] where x i is each data point in this dimension, and N is the total number of data points in this dimension.

[0112] Subsequently, each data point is normalized to obtain the normalized data S, such that the data in each dimension has zero mean and unit standard deviation. The normalization formula is as follows:

[0113] ;

[0114] Through the normalization process, it is ensured that the values of the three dimensions of the fault frequency, fault level, and impact range are in the same dimension, avoiding the dominant influence of a certain dimension on the diagnostic result.

[0115] Steps for weighted calculation of the fault mode correlation evaluation index:

[0116] Substitute the normalized multi-dimensional correlation index into the calculation formula of the fault mode correlation evaluation index FMAEI, which is used to measure the correlation strength between the abnormal phenomenon and the fault mode. The calculation formula is as follows:

[0117] ;

[0118] where α1, α2, and α3 are the weight coefficients of the three dimensions of the fault frequency, fault level, and impact range respectively, and the sum of the weight coefficients is 1; S f , Ss, S i are the data of the fault frequency, fault level, and impact range after normalization respectively.

[0119] It should be noted in this embodiment that in the process of converting explicit feedback into implicit feedback, the explicit feedback in the historical fault diagnosis information is processed using the multi-dimensional comprehensive index. In this way, the explicit feedback information is converted into implicit feedback through the multi-dimensional comprehensive index, thereby revealing the correlation strength between the abnormal phenomenon and the fault mode.

[0120] In one embodiment, in S4, a diagnostic recommendation model is constructed based on the RippleNet recommendation algorithm. This algorithm captures the complex correlation information in the equipment fault knowledge graph through a multi-order response propagation mechanism, and finally calculates the confirmation probability between the abnormal phenomenon and the fault mode.

[0121] In one embodiment, S4 includes the following steps:

[0122] S41: Calculate the correlation representation between abnormal phenomena and failure modes based on historical fault diagnosis data, and obtain the failure modes that meet the preset conditions historically for the input abnormal phenomena as seed failure modes; when specifically executing this step, consider the failure modes with strong correlation as seed failure modes.

[0123] S42: After inputting the abnormal phenomenon data, the failure modes with strong correlation in the historical diagnosis information will be used as the starting point of the model. Through the multi-level correlation relationships in the fault knowledge graph, the model gradually obtains the entities related to the abnormal phenomenon layer by layer through the iterative propagation mechanism, forming a ripple set. Based on the ripple set, obtain the failure modes in the propagation process, and jointly use the seed failure modes and the failure modes in the propagation process as candidate failure modes to participate in the subsequent calculation of the association probability. When specifically executing this step, starting from the abnormal phenomena that have occurred in the equipment history and the corresponding seed failure modes determined by diagnosis, obtain the relevant entities and ripple sets layer by layer through the links in the equipment fault knowledge graph.

[0124] S43: Calculate the association probability between the input abnormal phenomenon and the candidate failure modes based on the head entity and relationship of the ripple set.

[0125] S44: Cumulatively calculate the responses of each order of the ripple set layer by layer according to the association probability and the tail entity of the ripple set. This step simulates the process of fault propagation based on the connection information of the nodes in the graph, so as to obtain possible failure modes and related fault causes. The model accumulates the responses of each order, combines the corresponding candidate failure modes, and outputs the predicted diagnosis probability between the abnormal phenomenon and the failure mode. This probability reflects the diagnostic confidence level of diagnosing a certain abnormal phenomenon as a certain failure mode. When specifically executing this step, construct a prediction function to determine whether there is a potential diagnostic relationship between the abnormal phenomenon and the failure mode. The model form is:

[0126] ;

[0127] In the formula, represents the probability that the abnormal phenomenon w is diagnosed as the failure mode v, represents the model parameters of the function F.

[0128] In one embodiment, the calculation steps of the predicted diagnosis probability between the abnormal phenomenon and the candidate failure mode in S4 include:

[0129] S410: Obtain the 1-hop ripple set of the candidate failure mode v and the input abnormal phenomenon w in the equipment fault knowledge graph , by comparing the candidate failure mode v with the head entity h i and the relationship r i in the corresponding triple, for each triple (h i , r i, t i ) Assign an association probability p i :

[0130] ;

[0131] wherein, , is the embedding of the relationship r i and the head entity h i ; is the embedding of the candidate fault mode v, denotes the transposed embedding matrix of, d is the dimension of the embedding, p i represents the degree of correlation between v and h i , i is the number of triples in, Softmax is a normalization function that restricts the calculation results between 0 and 1 and ensures that the sum of all results is equal to 1.

[0132] S411: Take all the tail entities in and perform a weighted sum according to the corresponding association probability p i to obtain the diagnostic history V of the abnormal phenomenon w w for the first-order response vector of the candidate fault mode v :

[0133] ;

[0134] wherein, is the embedding of the tail entity t i ;

[0135] S412: Iteratively propagate the first-order response vector in the ripple set of the abnormal phenomenon w to obtain the high-order response of the abnormal phenomenon w , and the response of the abnormal phenomenon w to the candidate fault mode v is expressed as the sum of each order of response:

[0136] ;

[0137] wherein, is the H-th order response of the abnormal phenomenon w, and H is a positive integer greater than 2;

[0138] S413: Combine the abnormal phenomenon w and the candidate fault mode v to obtain the predicted probability:

[0139] ;

[0140] wherein, is the sigmoid function, is the embedding of the abnormal phenomenon w;

[0141] S414: Screen all candidate fault modes according to the predicted probability, obtain a recommended list of confirmed results, and provide a diagnostic reference for relevant technicians. In this embodiment, the top 5 recommended lists can be returned to assist relevant technicians in making diagnostic decisions.

[0142] In order to more intuitively demonstrate the use of the diagnosis recommendation model to perform equipment fault diagnosis recommendation tasks, the abnormal phenomenon of "insufficient power output" in the engine system during a certain task execution is used as an example to illustrate:

[0143] 1) When the abnormal phenomenon of "insufficient power output" is used as the diagnostic input, the model will automatically link to a series of fault modes strongly associated with this phenomenon based on historical diagnostic information, such as Figure 3 shown.

[0144] 2) After obtaining these strongly correlated fault modes, the model will further use the multi-level association relationships in the knowledge graph to conduct in-depth analysis and expansion of these modes. Taking the iteration of the fault mode "injector 1 fault" in the fault knowledge graph as an example to demonstrate its two-hop propagation, as shown in the figure below: Figure 4 shown.

[0145] This process mainly propagates in the following two directions: one direction is to combine the analysis of the root cause of the fault and find the fault mode that has a common causal relationship with the fault of injector 1 in the knowledge graph, such as the human factor "non-standard maintenance operation steps" and environmental factor "high temperature and high humidity environment corrosion" that cause the fault of injector 1, which will also cause "metering unit circuit break" and "abnormal wear of plunger". At the same time, in addition to this straightforward fault association, the diagnostic model is also good at excavating some hidden fault modes through the influence propagation between fault modes. For example, one of the reasons for the failure of injector 1 is the wear of the injection pump camshaft, and the wear of the injection pump camshaft will cause the injection pump to work unstably, which will cause the injection pressure to fluctuate, and finally induce the "injector 2 failure", thereby excavating deeper faults from the outside to the inside.

[0146] 3) After completing the multi-level association expansion of the 0-hop fault mode, the model will add a series of fault modes obtained during the propagation process to the candidate fault mode list. Then, based on the trained model parameters, the prediction probability of each candidate mode is calculated to evaluate the possibility of its existence of a diagnostic relationship under the current abnormal phenomenon. Finally, the system sorts the candidate fault modes according to these prediction probabilities and returns a recommendation list of the top 5. The calculation results are shown in Table 4.

[0147] Table 4 Fault diagnosis recommendation results

[0148]

[0149] The above has introduced in detail the equipment fault diagnosis recommendation method based on the knowledge graph and multi-dimensional correlation indicators provided by the present invention. In this embodiment, specific examples are used to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

[0150] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined in this embodiment can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown in this embodiment, but will conform to the widest scope consistent with the principles and novel features disclosed in this embodiment.

Claims

1. A method for equipment fault diagnosis recommendation based on a knowledge graph and multi-dimensional correlation indicators, characterized in that It includes the following steps: S1: Based on the abnormal phenomena of equipment units and the fault logics of human factors, machine factors, and environmental factors, define the semantic types and semantic relationships in the equipment fault domain, and construct an ontology model of the equipment fault knowledge graph; S2: Obtain fault texts from equipment fault cases, and based on the semantic types and semantic relationships in the equipment fault knowledge graph ontology model, use a knowledge extraction algorithm to convert the fault texts into triple-structured data, and construct an equipment fault knowledge graph; the equipment fault cases include historical fault diagnosis data; S3: Construct multi-dimensional association indicators for the corresponding fault modes of abnormal phenomena, use the multi-dimensional association indicators to optimize the representation of the association between abnormal phenomena and fault modes, and convert the explicit feedback in historical fault diagnosis data into implicit feedback; S4: Receive the input abnormal phenomenon, and based on the association representation, obtain the fault modes whose association meets the preset conditions from the historical fault diagnosis data as seed fault modes; construct a diagnostic recommendation model, starting from the seed fault modes, perform iterative propagation in the fault knowledge graph to obtain a ripple set, and obtain the fault modes during the propagation process based on the ripple set, and use the seed fault modes and the fault modes during the propagation process as candidate fault modes; Calculate the association probability between the input abnormal phenomenon and the candidate fault modes based on the ripple set, accumulate the responses of each order based on the association probability, output the predicted diagnosis probability, and determine the diagnosis result.

2. The equipment fault diagnosis recommendation method based on a knowledge graph and multi-dimensional correlation indicators according to claim 1, wherein The equipment fault knowledge graph ontology model has a multi-level structure, including: The unit layer is used to describe the spatial positions and structural association information of equipment, systems, and components; The logic layer is used to describe fault manifestations and their mechanisms, and analyze fault causes, including any one or a combination of the following: human factors, machine factors, and environmental factors; The maintenance layer is used to define maintenance measures and solutions.

3. A method for recommending equipment fault diagnosis based on a knowledge graph and multi-dimensional correlation indicators according to claim 1, characterized in that, The semantic types in the equipment fault knowledge graph ontology model include any combination of the following: equipment name, system name, component name, fault mode, abnormal phenomenon, human factor, machine factor, or environmental factor; the semantic relationships include any combination of the following: composition, attribution, location, cause, inducement, manifestation, or take.

4. A method for recommending equipment fault diagnosis based on a knowledge graph and multi-dimensional correlation indexes according to claim 1, characterized in that The equipment fault cases in S2 include structured text information and unstructured text information; among them, The structured text information includes historical fault data; The unstructured text information includes professional technical documents, fault troubleshooting records, and maintenance manuals.

5. A method for recommending equipment fault diagnosis based on a knowledge graph and multi-dimensional correlation indicators according to claim 4, characterized in that, Use the pre-trained BERT-BiLSTM-CRF model to perform named entity recognition on the unstructured text information; use the pre-trained Chinese-BERT-wwm-ext model to extract the relationships between fault-related entities in the unstructured text information.

6. A method for recommending equipment fault diagnosis based on a knowledge graph and multi-dimensional correlation indicators according to claim 4, characterized in that S3 includes: Obtain the abnormal phenomena and the corresponding diagnosed fault modes, the fault frequencies, fault levels, and influence scopes corresponding to the fault modes from the historical fault diagnosis data as multi-dimensional association indicators; among them, The fault frequency is used to record the number of times the abnormal phenomenon is diagnosed as the corresponding fault mode; The fault level is used to evaluate the severity of a fault according to the impact degree of the fault mode; The influence scope is used to measure the impact of the fault mode on each component in the system.

7. A method for recommending equipment fault diagnosis based on a knowledge graph and multi-dimensional correlation indicators according to claim 6, characterized in that The steps of using the multi-dimensional correlation index to optimize the correlation representation between the abnormal phenomenon and the fault mode in S3 include: Normalize the multi-dimensional correlation index to the same dimension; Substitute the normalized multi-dimensional correlation index into the calculation formula of the fault mode correlation evaluation index FMAEI, which is used to measure the correlation strength between the abnormal phenomenon and the fault mode. The calculation formula is as follows: ； where α1, α2, and α3 are the weight coefficients for the three dimensions of failure frequency, failure level, and impact scope respectively, and the sum of the weight coefficients is 1; S f , Ss, S i are the failure frequency, failure level, and impact scope data after standardization respectively.

8. A method for recommending equipment fault diagnosis based on a knowledge graph and multi-dimensional correlation indicators according to claim 1, characterized in that In S4, a diagnostic recommendation model is constructed based on the RippleNet recommendation algorithm.

9. A method for recommending equipment fault diagnosis based on a knowledge graph and multi-dimensional correlation indicators according to claim 4, characterized in that, S4 includes the following steps: S41: Calculate the correlation representation between the abnormal phenomenon and the fault mode according to the historical fault diagnosis data, and obtain the fault mode that satisfies the preset conditions historically with the input abnormal phenomenon as the seed fault mode; S42: Put the seed fault mode into the equipment fault knowledge graph as the starting point, obtain the relevant ripple set through iterative propagation, obtain the fault mode in the propagation process according to the ripple set, and use the seed fault mode and the fault mode in the propagation process as the candidate fault modes together; S43: Calculate the association probability between the input abnormal phenomenon and the candidate fault mode according to the head entity and relationship of the ripple set; S44: Accumulate the responses of each order of the ripple set layer by layer according to the association probability, accumulate the responses of each order, combine the corresponding candidate fault modes, output the predicted diagnosis probability, and determine the diagnosis result according to the predicted diagnosis probability.

10. A method for recommending equipment fault diagnosis based on a knowledge graph and multi-dimensional correlation indicators according to claim 1, characterized in that, The calculation steps of the predicted diagnosis probability between the abnormal phenomenon and the candidate fault mode in S4 include: Obtain the 1-hop ripple set of the candidate fault mode v and the input abnormal phenomenon w in the equipment fault knowledge graph , by comparing the candidate fault mode v with the head entity h i and the relationship r in the corresponding triple i to assign an association probability p to each triple (h i , r i , t i ) in i : ； wherein, , is the embedding of relationship r i and head entity h i , and is the embedding of candidate failure mode v, d is the dimension of the embedding, and p i represents the degree of correlation between v and h i , i is the number of triples in, and softmax( ) is the softmax function; Take all the tail entities in and according to the corresponding association probability p i perform a weighted sum to obtain the diagnostic history V of the abnormal phenomenon w w for the first-order response vector of the candidate fault mode v : ； In the formula, is the embedding of the tail entity t i ; For the first-order response vector in the ripple set of the abnormal phenomenon w perform iterative propagation to obtain the high-order response of the abnormal phenomenon w , and the response of the abnormal phenomenon w to the candidate fault mode v is expressed as the sum of the responses of each order: ； In the formula, is the H-th order response of the abnormal phenomenon w, where H is a positive integer greater than 2; Combine the abnormal phenomenon w and the candidate fault mode v to obtain the predicted probability: ； In the formula, is the sigmoid function, is the embedding of the abnormal phenomenon w; Screen all candidate fault modes according to the predicted probability to obtain a recommended list of diagnosis results.

Images