Railway equipment fault report analysis method and device

Abstract

The invention provides a railway equipment fault report analysis method and device, and the method comprises the steps: collecting fault report data of railway equipment, and carrying out the data cleaning and preprocessing; a pre-trained BERT model is used for processing a text sequence of the fault report, the BERT model adopts a 12-layer encoder in a Transform architecture as a basic component, context information is extracted from a front text and a back text, a word vector in each sentence is generated, and named entity recognition is carried out; establishing a railway fault knowledge graph according to the identified named entities; and displaying the constructed railway equipment fault knowledge graph through visual representation, performing statistical analysis on fault data in combination with a visual form, and revealing distribution and trend of faults. The method can provide a scientific basis and a favorable tool for maintenance and fault prevention of the railway equipment, effectively improves the efficiency and accuracy of fault analysis of the railway equipment, and improves the safety and operation efficiency of a railway system.

Description

Technical Field

[0001] The present invention relates to the field of railway technology; in particular, to railway equipment fault report analysis based on text data mining; specifically, to a railway equipment fault report analysis method and device. Background Art

[0002] With the rapid expansion of the railway network and the continuous increase in operating mileage, the railway department has accumulated a large number of fault reports related to railway equipment. These fault reports comprehensively record information such as fault time, location, cause and remedial measures, and are a comprehensive data source for analyzing failure modes and predicting future failures.

[0003] However, since these data are in long text format, it is difficult for managers to quickly and effectively understand and discover new knowledge hidden in the data.

[0004] Data mining is a technology that extracts valuable information and knowledge from massive data, covering multiple fields such as statistics, natural language processing, and machine learning. Through data mining, various things and phenomena can be deeply analyzed and understood, providing new possibilities for research and decision-making. The process of data mining usually includes steps such as data collection, preprocessing, structured transformation, analysis, result visualization, and knowledge discovery. Data mining technology can not only improve the efficiency of information processing, but also enhance insight and prediction capabilities. It is widely used in rail transportation, scientific research, medical diagnosis and other fields.

[0005] At present, a large amount of document data has been generated in the process of railway operation. In response to these rich documents, many research institutions and scholars use text data mining technology to conduct in-depth analysis, aiming to identify the occurrence patterns and characteristics of events, faults or accidents. These studies provide managers with more comprehensive and detailed decision support by extracting and analyzing key information in the text. Hou Tong et al. "A method for analyzing the causes of fine-grained faults in track circuits based on text mining. Journal of the China Railway Society 2022; 44 (10): 73-81" uses support vector machines and K-means algorithms to conduct text mining on the causes of fine-grained faults in track circuits. Lin Haixiang et al. "An automatic classification method for railway signal equipment faults based on text mining. Journal of Yunnan University (Natural Science Edition) 2022; 44 (02)" combines Word2vec, SMOTE algorithm and convolutional neural network to realize an automatic classification method for railway signal equipment fault text. Xu Wei et al. "Construction and analysis of railway accident causal network based on association rules. Railway Transportation and Economy 2020; 42 (11): 72-79" applied Apriori algorithm to mine railway accident causal association rules, constructed railway accident causal network, and analyzed the community structure and topological characteristics of the network. Tian Shirun et al. "Compound turnout fault diagnosis based on Bayesian network. Journal of Beijing Jiaotong University 2020; 44 (06): 118-125" In view of the lack of targetedness in the current daily maintenance strategy of compound turnouts, it is proposed to use text mining technology to diagnose compound turnout faults.

[0006] However, railway fault text data is characterized by its complexity, integrating entities such as fault location, fault type, fault time, and fault category, showing more diverse characteristics compared with ordinary text data. Fault-related entities are often more lengthy and contain a wide range of railway-specific terms, such as detailed fault descriptions, and the complexity is more obvious. In addition, the scarcity of publicly accessible datasets in the field of railway technology also hinders the identification of various types of railway equipment fault knowledge in the text data mining process. Summary of the invention

[0007] In view of this, the purpose of the present invention is to propose a railway equipment fault report analysis method and device, which realizes systematic analysis and visual representation of railway faults through named entity recognition and knowledge graph construction; uses BERT model and BiLSTM module for semantic encoding, and decodes through CRF layer to identify key entities in text, and constructs knowledge graph to intuitively display fault information, which is convenient for users to perform interactive query and analysis; by combining advanced natural language processing technology and graph database technology, the efficiency and accuracy of railway equipment fault analysis are improved, and a scientific basis and favorable tools are provided for the maintenance and fault prevention of railway equipment, thereby improving the safety and operation efficiency of the railway system.

[0008] The present invention provides a railway equipment fault report analysis method, comprising the following steps:

[0009] S1. Collecting fault report data of railway equipment, cleaning and preprocessing the fault report data, and laying a foundation for subsequent fault analysis;

[0010] S2. Use the pre-trained BERT model to process the text sequence of the fault report data after preprocessing. The BERT model uses the 12-layer encoder in the Transformer architecture as the basic component to extract context information from the previous and next texts, generate word vectors in each sentence, and perform named entity recognition;

[0011] S3. Establish a railway fault knowledge graph based on the identified named entities;

[0012] The construction of the knowledge graph helps to fully understand the causes and patterns of failures, and provide support for the maintenance and failure prevention of railway equipment.

[0013] S4. Through visual representation, the constructed railway equipment fault knowledge graph is displayed. Combined with the visualization form, the fault data is statistically analyzed to reveal the distribution and trend of the faults.

[0014] Through visualization tools, the information in the knowledge graph is intuitively displayed, which is convenient for users to conduct interactive queries and analysis; combined with visualization forms, statistical analysis of fault data is performed to reveal the distribution and trends of faults and provide decision support.

[0015] Furthermore, the step S2 includes the following steps:

[0016] S21, process the character sequence input by the model input layer, perform word segmentation according to the predetermined maximum sequence length, add the [CLS] entity tag at the beginning of the sequence, and add the [SEP] entity tag between two sentences; use the embedding of the three vectors of Token embedding, Segment embedding and Position embedding to generate the input sequence word vector;

[0017] S22. The semantic features of each input word vector are extracted through a bidirectional Transformer, and a vector representation with rich semantic features is extracted for each input word. Based on the Transformer attention mechanism, the weight coefficient of the word is dynamically adjusted according to the correlation between the words in the sentence to obtain the vector representation of the word in each sentence. The expression of the vector representation of the word is: (1)

[0018] In formula (1), Q, K and V represent word embedding matrices, dk represents the embedding dimension;

[0019] Based on the multi-head attention mechanism, the word vector matrices Q, K and V are mapped to multiple different linear transformations, and different attentions are spliced ​​together;

[0020] S23, the BERT layer generates a word vector sequence by using the vector representation of the word as input to the BiLSTM layer for semantic encoding; the BiLSTM layer extracts richer context information by processing the input word vector sequence in both forward and backward directions;

[0021] The BiLSTM layer is an extended architecture of two independent LSTM layers that process the input word vector sequence in both the forward and backward directions. This bidirectional processing is particularly effective for tasks that require context of contextual elements to achieve accurate predictions.

[0022] S24, obtaining the outputs of the BERT layer and the BiLSTM layer, connecting the outputs of the BERT layer and the BiLSTM layer through a vector concatenation layer, extracting the deep features of the entity label of step S21 through a self-attention mechanism, and inputting the extracted deep features into a fully connected layer for dimensionality reduction processing, thereby simplifying the data for subsequent analysis and encoding and improving processing efficiency;

[0023] The self-attention mechanism can focus on pre-identified entity tags and strengthen the correlation and information exchange between entities and between tags within entities through weighted calculation.

[0024] S25, use the CRF layer to decode the deep feature information after dimensionality reduction, and calculate the probability of the state sequence Identify entities;

[0025] Given a specific text sequence And the corresponding label sequence , the probability of the state sequence is: (2)

[0026] In formula (2), is the normalization factor, is the state characteristic function; The calculation expression is: (3)

[0027] In formula (3), represents the corresponding association weight;

[0028] By identifying the tag sequence that achieves the maximum probability, the result S of entity recognition is determined* :

[0029] (4)

[0030] As one of the important components in the present invention, the CRF layer generates the optimal sequence of predicted labels to effectively represent the entities identified in the text.

[0031] Furthermore, the S3 step includes the following steps:

[0032] S31. Structural storage of the identified fault time, line, train number, fault description, cause analysis, treatment measures, fault location, category, responsible system and entities affected by the fault;

[0033] S32. Define the relationship between various entities, and use the neo4j graph database to visualize different entities and their relationships to form a railway fault knowledge graph.

[0034] Furthermore, the data cleaning method in step S1 includes:

[0035] Remove noise information and interfering data, such as duplicate records and invalid characters, to ensure data quality;

[0036] Unify data formats and standardize formats; for example, unify the date format to "YYYY-MM-DD-YYY" to ensure a unified data structure.

[0037] Normalize synonyms and spelling errors in the text to ensure uniform representation of the same entity;

[0038] Use word segmentation tools and BIO annotation methods to annotate the processed fault report data, and manually verify the word segmentation and part-of-speech tagging results to ensure the correct processing of railway fault-related terms and proper nouns;

[0039] The annotated text is vectorized. For example, the text id corresponding to "column tail host..." is "0123...", and the label id is {'B-LOC': '0', 'I-LOC': '1', 'I-LOC': '2', 'I-LOC': '3'....}.

[0040] Furthermore, the method for removing noise information and interference data includes: using regular expressions and data cleaning tools to automatically filter invalid data.

[0041] The present invention also provides a railway equipment fault report analysis device, which executes the railway equipment fault report analysis method as described above, comprising:

[0042] Data acquisition and preprocessing module: used to collect fault report data of railway equipment, perform data cleaning and preprocessing on the fault report data, and lay the foundation for subsequent analysis;

[0043] Equipment fault report analysis module: including:

[0044] Named Entity Recognition submodule: It is used to process the text sequence of the fault report data after preprocessing using the pretrained BERT model. The BERT model uses the 12-layer encoder in the Transformer architecture as the basic component to extract context information from the previous and next texts, generate word vectors in each sentence, and perform named entity recognition.

[0045] Railway fault knowledge graph construction submodule: used to build a railway fault knowledge graph based on the identified named entities;

[0046] Visualization and analysis module: It is used to display the constructed railway equipment fault knowledge graph through visual representation, and to conduct statistical analysis on fault data in combination with visualization to reveal the distribution and trend of faults.

[0047] Furthermore, the named entity recognition submodule includes:

[0048] Entity tagging unit: used to process the character sequence input by the model input layer, perform word segmentation according to the predetermined maximum sequence length, add the [CLS] entity tag at the beginning of the sequence, and add the [SEP] entity tag between two sentences; generate the input sequence word vector by embedding the three vectors of Token embedding, Segment embedding and Position embedding;

[0049] The entity extraction unit includes:

[0050] Extract vector representation subunit: It is used to extract the semantic features of each input word vector through a bidirectional Transformer, and extract a vector representation with rich semantic features for each input word; based on the Transformer attention mechanism, the weight coefficient of the word is dynamically adjusted according to the correlation between the words in the sentence to obtain the vector representation of the word in each sentence; based on the multi-head attention mechanism, the word vector matrices Q, K and V are mapped to multiple different linear transformations, and different attentions are spliced ​​together;

[0051] Feature extraction subunit: used for providing the word vector sequence generated by the BERT layer from the vector representation of the word as input to the BiLSTM layer for semantic encoding; the BiLSTM layer extracts richer context information by processing the input word vector sequence in both forward and backward directions;

[0052] Deep feature extraction subunit: used to obtain the output of the BERT layer and the BiLSTM layer, connect the output of the BERT layer and the BiLSTM layer through the vector concatenation layer, extract the deep features of the entity label of step S21 through the self-attention mechanism, and input the extracted deep features into the fully connected layer for dimensionality reduction processing;

[0053] Sequence decoding subunit: It is used to decode the deep feature information after dimensionality reduction using the CRF layer, identify entities by calculating the probability of the state sequence, and determine the result of entity recognition by identifying the label sequence with the maximum probability.

[0054] Furthermore, the railway fault knowledge graph construction submodule includes:

[0055] Entity structured storage unit: used to structure the storage of the identified fault time, line, train number, fault description, cause analysis, treatment measures, fault location, category, responsible system and entities affected by the fault;

[0056] Relationship definition and storage unit: used to define the relationship between various entities, and use the neo4j graph database to visualize different entities and their relationships to form a railway fault knowledge graph.

[0057] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the railway equipment fault report analysis method as described above are implemented.

[0058] The present invention also provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the railway equipment fault report analysis method as described above when executing the program.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] The railway equipment fault report analysis method and device provided by the present invention realize systematic analysis and visual representation of railway faults through named entity recognition and knowledge graph construction, use BERT model and BiLSTM module for semantic encoding, and decode through CRF layer to identify key entities in the text. The constructed knowledge graph can intuitively display fault information, which is convenient for users to perform interactive query and analysis; combined with advanced natural language processing technology and graph database technology, it can provide scientific basis and favorable tools for maintenance and fault prevention of railway equipment, effectively improve the efficiency and accuracy of railway equipment fault analysis, and improve the safety and operation efficiency of railway system. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.

[0062] In the attached picture:

[0063] Figure 1 A flow chart of a railway equipment fault report analysis method provided by an embodiment of the present invention;

[0064] Figure 2 A schematic diagram of railway equipment fault report data preprocessing provided by an embodiment of the present invention;

[0065] Figure 3 A schematic diagram of the architecture of named entity recognition for railway equipment fault report analysis provided by an embodiment of the present invention;

[0066] Figure 4 A schematic diagram of entity relationships provided by an embodiment of the present invention;

[0067] Figure 5 A schematic diagram of the structure of a railway equipment fault report analysis device provided by an embodiment of the present invention;

[0068] Figure 6 A flow chart of a railway equipment fault report analysis method according to the present invention;

[0069] Figure 7 It is a flow chart of step S2 of the railway equipment fault report analysis method of the present invention;

[0070] Figure 8 It is a flow chart of step S3 of the railway equipment fault report analysis method of the present invention;

[0071] Fig. 9 The figure is a schematic diagram of the structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0072] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and products consistent with some aspects of the present disclosure as detailed in the appended claims.

[0073] The terms used in this disclosure are for the purpose of describing specific embodiments only and are not intended to limit the disclosure. The singular forms of "a", "said" and "the" used in this disclosure and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0074] It should be understood that although the terms first, second, third, etc. may be used in the present disclosure to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present disclosure, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0075] The embodiments of the present invention are described in further detail below.

[0076] The present invention provides a railway equipment fault report analysis method, see Figure 6 As shown, the following steps are included:

[0077] S1. Collecting fault report data of railway equipment, and performing data cleaning and preprocessing on the fault report data (such as Figure 2 As shown), laying the foundation for subsequent fault analysis;

[0078] The data cleaning method comprises:

[0079] Remove noise information and interfering data, including duplicate records, invalid characters, etc. Use regular expressions and data cleaning tools to automatically filter invalid data to ensure data quality.

[0080] Unify data formats and standardize formats; for example, unify the date format to "YYYY-MM-DD-YYY" to ensure a unified data structure.

[0081] Normalize synonyms and spelling errors in the text to ensure uniform representation of the same entity;

[0082] Use word segmentation tools and BIO annotation methods to annotate the processed fault report data, and manually verify the word segmentation and part-of-speech tagging results to ensure the correct processing of railway fault-related terms and proper nouns;

[0083] Perform text vectorization on the annotated text.

[0084] In this embodiment, the text id corresponding to "column tail host..." is "0123...", and the tag id is {'B-LOC':'0', 'I-LOC':'1', 'I-LOC':'2', 'I-LOC':'3'....}.

[0085] According to the railway fault knowledge structure, the embodiment of the present invention defines 10 types of entities and 9 types of entity relationships, as shown in Table 1.

[0086] Table 1

[0087] Preferably, the processed data in Table 1 is stored in a NoSQL database to facilitate subsequent processing and analysis. Distributed storage technology is used to ensure the scalability and high availability of data.

[0088] S2. Use the pre-trained BERT model (such as Figure 3 The BERT model uses a 12-layer encoder in the Transformer architecture as a basic component to extract context information from the previous and next texts, generate word vectors in each sentence, and perform named entity recognition. Specifically, see Figure 7 As shown, the following steps are included:

[0089] S21, process the character sequence input by the model input layer, perform word segmentation according to the predetermined maximum sequence length, add the [CLS] entity tag at the beginning of the sequence, and add the [SEP] entity tag between two sentences; use the embedding of the three vectors of Token embedding, Segment embedding and Position embedding to generate the input sequence word vector;

[0090] S22. The semantic features of each input word vector are extracted through a bidirectional Transformer, and a vector representation with rich semantic features is extracted for each input word. Based on the Transformer attention mechanism, the weight coefficient of the word is dynamically adjusted according to the correlation between the words in the sentence to obtain the vector representation of the word in each sentence. The expression of the vector representation of the word is:

[0091] (1)

[0092] In formula (1), Q, K and V represent word embedding matrices, d k represents the embedding dimension;

[0093] Based on the multi-head attention mechanism, the word vector matrices Q, K and V are mapped to multiple different linear transformations, and different attentions are spliced ​​together;

[0094] S23, the BERT layer generates a word vector sequence by using the vector representation of the word as input to the BiLSTM layer for semantic encoding; the BiLSTM layer extracts richer context information by processing the input word vector sequence in both forward and backward directions;

[0095] The BiLSTM layer is an extended architecture of two independent LSTM layers that process the input word vector sequence in both the forward and backward directions. This bidirectional processing is particularly effective for tasks that require context of contextual elements to achieve accurate predictions.

[0096] S24, obtaining the outputs of the BERT layer and the BiLSTM layer, connecting the outputs of the BERT layer and the BiLSTM layer through a vector concatenation layer, extracting the deep features of the entity label of step S21 through a self-attention mechanism, and inputting the extracted deep features into a fully connected layer for dimensionality reduction processing, thereby simplifying the data for subsequent analysis and encoding and improving processing efficiency;

[0097] The self-attention mechanism can focus on pre-identified entity tags and strengthen the correlation and information exchange between entities and between tags within entities through weighted calculation.

[0098] S25, use the CRF layer to decode the deep feature information after dimensionality reduction, and calculate the probability of the state sequence Identify entities;

[0099] Given a specific text sequence And the corresponding label sequence , the probability of the state sequence is:

[0100] (2)

[0101] In formula (2), is the normalization factor, is the state characteristic function; The calculation expression is:

[0102] (3)

[0103] In formula (3), represents the corresponding association weight;

[0104] By identifying the tag sequence that achieves the maximum probability, the result S of entity recognition is determined * :

[0105] (4)

[0106] As one of the important components in the present invention, the CRF layer generates the optimal sequence of predicted labels to effectively represent the entities identified in the text.

[0107] S3. Based on the identified named entities, a railway fault knowledge graph is established; specifically, see Figure 8 As shown, the following steps are included:

[0108] S31. Structural storage of the identified fault time, line, train number, fault description, cause analysis, treatment measures, fault location, category, responsible system and entities affected by the fault;

[0109] S32. Define the relationships between various entities (such as Figure 4 The different entities and their relationships are visualized using the neo4j graph database to form a railway fault knowledge graph.

[0110] The construction of the knowledge graph helps to fully understand the causes and patterns of failures, and provide support for the maintenance and failure prevention of railway equipment.

[0111] S4. Through visual representation, the constructed railway equipment fault knowledge graph is displayed. Combined with the visualization form, the fault data is statistically analyzed to reveal the distribution and trend of the faults.

[0112] Through visualization tools, the information in the knowledge graph is intuitively displayed, which is convenient for users to conduct interactive queries and analysis; combined with visualization forms, statistical analysis of fault data is performed to reveal the distribution and trends of faults and provide decision support.

[0113] Figure 1 The basic process of the railway equipment fault report analysis method of this embodiment is shown.

[0114] The embodiment of the present invention also provides a railway equipment fault report analysis device, which executes the railway equipment fault report analysis method as described above, see Figure 5 As shown, including:

[0115] Data acquisition and preprocessing module: used to collect fault report data of railway equipment, perform data cleaning and preprocessing on the fault report data, and lay the foundation for subsequent analysis;

[0116] Equipment fault report analysis module: including:

[0117] Named Entity Recognition submodule: It is used to process the text sequence of the fault report data after preprocessing using the pretrained BERT model. The BERT model uses the 12-layer encoder in the Transformer architecture as the basic component to extract context information from the previous and next texts, generate word vectors in each sentence, and perform named entity recognition.

[0118] Railway fault knowledge graph construction submodule: used to build a railway fault knowledge graph based on the identified named entities;

[0119] Visualization and analysis module: It is used to display the constructed railway equipment fault knowledge graph through visual representation, and to conduct statistical analysis on fault data in combination with visualization to reveal the distribution and trend of faults.

[0120] The named entity recognition submodule includes:

[0121] Entity tagging unit: used to process the character sequence input by the model input layer, perform word segmentation according to the predetermined maximum sequence length, add the [CLS] entity tag at the beginning of the sequence, and add the [SEP] entity tag between two sentences; generate the input sequence word vector by embedding the three vectors of Token embedding, Segment embedding and Position embedding;

[0122] The entity extraction unit includes:

[0123] Extract vector representation subunit: It is used to extract the semantic features of each input word vector through a bidirectional Transformer, and extract a vector representation with rich semantic features for each input word; based on the Transformer attention mechanism, the weight coefficient of the word is dynamically adjusted according to the correlation between the words in the sentence to obtain the vector representation of the word in each sentence; based on the multi-head attention mechanism, the word vector matrices Q, K and V are mapped to multiple different linear transformations, and different attentions are spliced ​​together;

[0124] Feature extraction subunit: used for providing the word vector sequence generated by the BERT layer from the vector representation of the word as input to the BiLSTM layer for semantic encoding; the BiLSTM layer extracts richer context information by processing the input word vector sequence in both forward and backward directions;

[0125] Deep feature extraction subunit: used to obtain the output of the BERT layer and the BiLSTM layer, connect the output of the BERT layer and the BiLSTM layer through the vector concatenation layer, extract the deep features of the entity label of step S21 through the self-attention mechanism, and input the extracted deep features into the fully connected layer for dimensionality reduction processing;

[0126] Sequence decoding subunit: It is used to decode the deep feature information after dimensionality reduction using the CRF layer, identify entities by calculating the probability of the state sequence, and determine the result of entity recognition by identifying the label sequence with the maximum probability.

[0127] The railway fault knowledge graph construction submodule includes:

[0128] Entity structured storage unit: used to structure the storage of the identified fault time, line, train number, fault description, cause analysis, treatment measures, fault location, category, responsible system and entities affected by the fault;

[0129] Relationship definition and storage unit: used to define the relationship between various entities, and use the neo4j graph database to visualize different entities and their relationships to form a railway fault knowledge graph.

[0130] An embodiment of the present invention further provides a computer device, Fig. 9 is a schematic diagram of the structure of a computer device provided by an embodiment of the present invention; see the accompanying drawings Fig. 9 As shown, the computer device includes: an input device 23, an output device 24, a memory 22 and a processor 21; the memory 22 is used to store one or more programs; when the one or more programs are executed by the one or more processors 21, the one or more processors 21 implement the railway equipment fault report analysis method provided in the above embodiment; wherein the input device 23, the output device 24, the memory 22 and the processor 21 can be connected by a bus or other means, Fig. 9 The example of connecting through bus is taken in the following.

[0131] The memory 22 is a readable and writable storage medium of a computing device, which can be used to store software programs and computer executable programs, such as program instructions corresponding to the railway equipment fault report analysis method described in the embodiment of the present invention; the memory 22 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required for a function; the data storage area can store data created according to the use of the device, etc.; in addition, the memory 22 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device; in some instances, the memory 22 can further include a memory remotely arranged relative to the processor 21, and these remote memories can be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0132] The input device 23 may be used to receive input digital or character information, and to generate key signal input related to user settings and function control of the device; the output device 24 may include a display device such as a display screen.

[0133] The processor 21 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 22, that is, realizes the above-mentioned railway equipment fault report analysis method.

[0134] The computer device provided above can be used to execute the railway equipment fault report analysis method provided in the above embodiment, and has corresponding functions and beneficial effects.

[0135] The embodiment of the present invention also provides a storage medium containing computer executable instructions, which are used to execute the railway equipment fault report analysis method provided in the above embodiment when executed by a computer processor. The storage medium is any of various types of memory devices or storage devices, and the storage medium includes: installation media, such as CD-ROM, floppy disk or tape device; computer system memory or random access memory, such as DRAM, DDR RAM, SRAM, EDORAM, Rambus RAM, etc.; non-volatile memory, such as flash memory, magnetic media (such as hard disk or optical storage); registers or other similar types of memory elements, etc.; the storage medium may also include other types of memory or a combination thereof; in addition, the storage medium may be located in a first computer system in which the program is executed, or may be located in a different second computer system, which is connected to the first computer system via a network (such as the Internet); the second computer system may provide program instructions to the first computer for execution. The storage medium includes two or more storage media that may reside in different locations (for example, in different computer systems connected via a network). The storage medium may store program instructions (for example, specifically implemented as a computer program) that can be executed by one or more processors.

[0136] Of course, the storage medium containing computer executable instructions provided in an embodiment of the present invention is not limited to the railway equipment fault report analysis method described in the above embodiment, and can also execute related operations in the railway equipment fault report analysis method provided in any embodiment of the present invention.

[0137] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0138] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A railway equipment fault report analysis method, characterized in that: The following steps are involved: S1. Collecting fault report data of railway equipment, cleaning and preprocessing the fault report data, and laying a foundation for subsequent fault analysis; S2. Use the pre-trained BERT model to process the text sequence of the fault report data after preprocessing. The BERT model uses the 12-layer encoder in the Transformer architecture as the basic component to extract context information from the previous and next texts, generate word vectors in each sentence, and perform named entity recognition; S3. Establish a railway fault knowledge graph based on the identified named entities; S4. Through visual representation, the constructed railway equipment fault knowledge graph is displayed. Combined with the visualization form, the fault data is statistically analyzed to reveal the distribution and trend of the faults.

2. The railway equipment fault report analysis method according to claim 1, characterized in that: The S2 step includes the following steps: S21, process the character sequence input by the model input layer, perform word segmentation according to the predetermined maximum sequence length, add the [CLS] entity tag at the beginning of the sequence, and add the [SEP] entity tag between two sentences; use the embedding of the three vectors of Token embedding, Segment embedding and Position embedding to generate the input sequence word vector; S22. The semantic features of each input word vector are extracted through a bidirectional Transformer, and a vector representation with rich semantic features is extracted for each input word. Based on the Transformer attention mechanism, the weight coefficient of the word is dynamically adjusted according to the correlation between the words in the sentence to obtain the vector representation of the word in each sentence. The expression of the vector representation of the word is: （1） In formula (1), Q, K and V represent word embedding matrices, d k represents the embedding dimension; Based on the multi-head attention mechanism, the word vector matrices Q, K and V are mapped to multiple different linear transformations, and different attentions are spliced ​​together; S23, the BERT layer generates a word vector sequence by using the vector representation of the word as input to the BiLSTM layer for semantic encoding; the BiLSTM layer extracts richer context information by processing the input word vector sequence in both forward and backward directions; S24, obtaining the outputs of the BERT layer and the BiLSTM layer, connecting the outputs of the BERT layer and the BiLSTM layer through a vector concatenation layer, extracting the deep features of the entity label of step S21 through a self-attention mechanism, and inputting the extracted deep features into a fully connected layer for dimensionality reduction processing; S25, use the CRF layer to decode the deep feature information after dimensionality reduction, and calculate the probability of the state sequence Identify entities; Given a specific text sequence And the corresponding label sequence , the probability of the state sequence is: （2） In formula (2), is the normalization factor, is the state characteristic function; The calculation expression is: （3） In formula (3), represents the corresponding association weight; By identifying the tag sequence that achieves the maximum probability, the result S of entity recognition is determined * : （4）。 3. The railway equipment fault report analysis method according to claim 1, characterized in that: The S3 step includes the following steps: S31. Structural storage of the identified fault time, line, train number, fault description, cause analysis, treatment measures, fault location, category, responsible system and entities affected by the fault; S32. Define the relationship between various entities, and use the neo4j graph database to visualize different entities and their relationships to form a railway fault knowledge graph.

4. The railway equipment fault report analysis method according to claim 1, characterized in that: The data cleaning method in step S1 includes: Remove noise information and interference data; Unify data formats and standardize formats; Normalize synonyms and spelling errors in the text to ensure uniform representation of the same entity; Use word segmentation tools and BIO annotation methods to annotate the processed fault report data, and manually verify the word segmentation and part-of-speech tagging results to ensure the correct processing of railway fault-related terms and proper nouns; Perform text vectorization on the annotated text.

5. The railway equipment fault report analysis method according to claim 4, characterized in that: The method for removing noise information and interference data includes: using regular expressions and data cleaning tools to automatically filter invalid data.

6. A railway equipment fault report analysis device, executing the railway equipment fault report analysis method according to any one of claims 1 to 5, characterized in that: include: Data acquisition and preprocessing module: used to collect fault report data of railway equipment, perform data cleaning and preprocessing on the fault report data, and lay the foundation for subsequent analysis; Equipment fault report analysis module: including: Named Entity Recognition submodule: It is used to process the text sequence of the fault report data after preprocessing using the pretrained BERT model. The BERT model uses the 12-layer encoder in the Transformer architecture as the basic component to extract context information from the previous and next texts, generate word vectors in each sentence, and perform named entity recognition. Railway fault knowledge graph construction submodule: used to build a railway fault knowledge graph based on the identified named entities; Visualization and analysis module: It is used to display the constructed railway equipment fault knowledge graph through visual representation, and to conduct statistical analysis on fault data in combination with visualization to reveal the distribution and trend of faults.

7. The railway equipment fault report analysis device according to claim 6, characterized in that: The named entity recognition submodule includes: Entity tagging unit: used to process the character sequence input by the model input layer, perform word segmentation according to the predetermined maximum sequence length, add the [CLS] entity tag at the beginning of the sequence, and add the [SEP] entity tag between two sentences; generate the input sequence word vector by embedding the three vectors of Token embedding, Segment embedding and Position embedding; The entity extraction unit includes: Extract vector representation subunit: It is used to extract the semantic features of each input word vector through a bidirectional Transformer, and extract a vector representation with rich semantic features for each input word; based on the Transformer attention mechanism, the weight coefficient of the word is dynamically adjusted according to the correlation between the words in the sentence to obtain the vector representation of the word in each sentence; based on the multi-head attention mechanism, the word vector matrices Q, K and V are mapped to multiple different linear transformations, and different attentions are spliced ​​together; Feature extraction subunit: used for providing the word vector sequence generated by the BERT layer from the vector representation of the word as input to the BiLSTM layer for semantic encoding; the BiLSTM layer extracts richer context information by processing the input word vector sequence in both forward and backward directions; Deep feature extraction subunit: used to obtain the output of the BERT layer and the BiLSTM layer, connect the output of the BERT layer and the BiLSTM layer through the vector concatenation layer, extract the deep features of the entity label of step S21 through the self-attention mechanism, and input the extracted deep features into the fully connected layer for dimensionality reduction processing; Sequence decoding subunit: It is used to decode the deep feature information after dimensionality reduction using the CRF layer, identify entities by calculating the probability of the state sequence, and determine the result of entity recognition by identifying the label sequence with the maximum probability.

8. The railway equipment fault report analysis device according to claim 6, characterized in that: The railway fault knowledge graph construction submodule includes: Entity structured storage unit: used to structure the storage of the identified fault time, line, train number, fault description, cause analysis, treatment measures, fault location, category, responsible system and entities affected by the fault; Relationship definition and storage unit: used to define the relationship between various entities, and use the neo4j graph database to visualize different entities and their relationships to form a railway fault knowledge graph.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the railway equipment fault report analysis method according to any one of claims 1 to 5 are implemented.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the railway equipment fault report analysis method as described in any one of claims 1 to 5 are implemented.

Images