Railway green performance data knowledge extraction method

Abstract

The invention provides a railway green performance data knowledge extraction method. The method comprises the steps of data collection and preprocessing, knowledge graph construction and data standardized storage. The method comprises the following steps: collecting and classifying key data of a full life cycle of a railway; constructing an ontology by a top-down method, extracting knowledge, and storing the knowledge in a graph database; storing the structured data based on the metadata, sorting and classifying the unstructured data, binding the unstructured data with the knowledge graph, and storing the unstructured data; according to the method, the environment-friendly degree of railway engineering can be improved, information management is optimized, the data driving force of decision support is increased, data sharing and communication are promoted, the sustainability of railway projects is improved, compliance of environmental protection rules and policies is achieved, and the timeliness and dynamic updating capacity of data are improved; not only are knowledge extraction efficiency and accuracy improved, but also monitoring and improvement of key performance indexes of the railway engineering are more scientific and systematic, and energy conservation, consumption reduction and ecological environmental protection of the railway engineering are further promoted.

Description

Technical Field

[0001] The invention relates to the technical field of data processing and environmental engineering, and in particular to a method for extracting knowledge from railway green performance data. Background Art

[0002] Railway green performance data involves the increasing emphasis on environmental protection and sustainable development in railway construction and operation while focusing on economic benefits; railway green performance data focuses on the environmental friendliness of the railway system in various links such as design, construction, maintenance and operation, including various indicators such as energy consumption, emissions, resource utilization, and ecological impact;

[0003] In the traditional process of extracting knowledge from railway green performance data, most of the data is collected and sorted manually, relying on paper records and simple spreadsheets. With the improvement of big data technology and information technology, the railway system has begun to try to use more advanced data collection and analysis tools, but faces the following defects: 1. Data island problem. Data in different departments and systems are isolated from each other, lacking effective data integration and sharing mechanisms, resulting in low data utilization; 2. Outdated analysis methods. Early data processing and analysis methods cannot handle large-scale and highly complex data sets well, and the accuracy and operability of analysis results are limited. 3. Insufficient automation. In the existing railway green performance data collection and analysis process, many steps still require manual intervention, which is inefficient and prone to errors. 4. Knowledge extraction is not in-depth. Traditional data analysis methods usually stay on the surface, lack deep knowledge extraction, and lack understanding of the patterns and relationships behind the data; 5. Poor real-time performance. The real-time performance of monitoring and feedback of environmental data is insufficient, which affects the ability to respond and adjust quickly; 6. Weak fusion ability. The ability to effectively integrate data of different types and sources is insufficient, and the information contained in the data cannot be fully evaluated and utilized.

[0004] Therefore, the present application provides a railway green performance data knowledge extraction method to solve the above problems. Summary of the invention

[0005] The present invention provides a method for extracting knowledge from railway green performance data, which aims to fully extract the knowledge content contained in the "four savings and one environmental protection" data in railway engineering, thereby supporting the decision-making and management of railway engineering design, construction and operation.

[0006] The present invention provides a method for extracting knowledge of railway green performance data, comprising:

[0007] Step 1: Data collection and preprocessing: collect key data of “four conservations and one environmental protection” in the railway design, construction and operation stages, improve the collected key data, and conduct preliminary classification;

[0008] Step 2: Knowledge graph construction. The ontology of the entire railway life cycle is constructed mainly using a top-down approach. Knowledge extraction operations are performed according to different data types. The relationship between the extraction results and the ontology is established and stored using a graph database. The repeated and conflicting information in heterogeneous data sources is integrated to generate a knowledge graph.

[0009] Step 3: In the later stage of building the railway green performance data knowledge graph, the railway life cycle ontology is automatically updated with data-driven;

[0010] Step 4: Standardize data and store it in the database. Use metadata-based data model construction and management technology to store structured data. For unstructured data, organize and classify the format and content of unstructured data based on various existing data processing algorithms and image processing algorithms. Then, bind the data to specific subject nodes one by one according to the data classification of the knowledge graph to complete the storage of unstructured data.

[0011] According to a railway green performance data knowledge extraction method provided by the present invention, the data collected in the data collection process are the key data of "four savings and one environmental protection" in railway engineering, specifically including energy-saving data, water-saving data, land-saving data, material-saving data and environmental protection data;

[0012] The above five types of data are collected once and regularly during the design, construction and operation stages of railway projects.

[0013] According to a railway green performance data knowledge extraction method provided by the present invention, the data preprocessing process includes:

[0014] Format the collected data and convert the files into a processable format; determine the data integrity and quality, and clean irrelevant or erroneous data;

[0015] Standardize the naming of data, clarify the names of specific data contact persons, and make preliminary classifications of data in the form of videos and pictures;

[0016] The preliminary classification process divides the data into structured data and unstructured data according to its structural form; the unstructured data also includes semi-structured data and electronic data. The main way to obtain unstructured data is through surveys conducted by construction units, design units, construction units, supervision units and railway bureau companies.

[0017] According to a railway green performance data knowledge extraction method provided by the present invention, the knowledge graph construction process includes:

[0018] The ontology construction mainly adopts the top-down method, using the data of railway design, construction and operation stages and the suggestions of experts in this field to establish the railway life cycle ontology;

[0019] Knowledge extraction: quality inspection of structured data and extraction based on rules; named entity recognition, relationship extraction, and attribute extraction are required for unstructured data to generate knowledge content for each data;

[0020] Knowledge fusion: store the knowledge content in the knowledge data framework, and perform data fusion when there is data duplication and conflict; associate the knowledge data framework with the railway life cycle ontology, improve and generate the entity relationship network, and perform quality assessment based on the current updated knowledge;

[0021] Generate a graph, and when the quality assessment result meets the preset conditions, generate a railway green performance data knowledge graph based on the entity relationship network.

[0022] According to a railway green performance data knowledge extraction method provided by the present invention, in the later stage of the railway green performance data knowledge graph construction, the railway full life cycle ontology is automatically updated with data-driven, and new ontology classifications will appear along with the updates of equipment, personnel, and data subjects, and the ontology is expanded through data-driven automatic updates;

[0023] The data-driven automatic update of the railway life cycle ontology includes the following processes:

[0024] Continuously monitor incoming data streams from various sensors, devices, and systems, automatically identifying and capturing new instances or changes;

[0025] Use machine learning and natural language processing algorithms to automatically extract new knowledge from real-time monitored data for expansion or modification of existing ontologies;

[0026] On top of automated updates, an expert review process is set up, where experts can manually review and adjust the ontology updates recommended by the algorithm.

[0027] Set up a data conflict mechanism to integrate or adjust the old ontology classification based on the access of new data sources;

[0028] Set up a relationship management mechanism to manage the relationship between historical data and new data to ensure a smooth transition between old and new data;

[0029] Set up reasoning mechanisms to infer new associations and insights from existing information.

[0030] According to a railway green performance data knowledge extraction method provided by the present invention, in the knowledge extraction process, the electronic data is extracted using a relevant algorithm;

[0031] Among them, paper materials, video images, and voice materials in electronic data are processed using text recognition, image recognition, and voice recognition algorithms to generate text, and then knowledge is extracted using text processing operations.

[0032] According to a railway green performance data knowledge extraction method provided by the present invention, in the knowledge extraction process, the extraction process for unstructured data includes:

[0033] Named entity recognition, where a named entity refers to an entity with a specific meaning in the text. The recognition process is to identify three major categories and seven subcategories in the text;

[0034] Relation extraction, determining the semantic relationship between two or more entities after identifying the named entities;

[0035] Attribute extraction, used to obtain specific attributes of an entity;

[0036] Knowledge integration combines the identified named entities, relations, and attributes to form knowledge content.

[0037] According to a railway green performance data knowledge extraction method provided by the present invention, the knowledge fusion process includes:

[0038] After the knowledge content is extracted, the knowledge content is stored in the knowledge data framework using a graph database, which is a graph of points and edges for users and a low-dimensional dense vector for computers;

[0039] For data of different modalities and sources, there will be certain conflicts and duplications. It is necessary to integrate knowledge according to different information sources, and complete and remove duplicates from the knowledge graph. Among them, knowledge integration methods include ontology alignment and entity alignment.

[0040] The ontology alignment is used to determine whether elements or structures in two knowledge data frameworks should match;

[0041] The entity alignment is used to map all entities in the knowledge content into the same vector space, and to achieve entity disambiguation by calculating the entity distance;

[0042] After the ontology alignment and the entity alignment process are completed, knowledge reasoning is performed on the alignment results, and the alignment results and reasoning results are associated with the railway life cycle ontology; the entity relationship network is generated and quality assessment is performed; if there is newly extracted knowledge content, the entity relationship network is updated and the quality assessment is repeated.

[0043] According to a railway green performance data knowledge extraction method provided by the present invention, the data standardization storage process includes:

[0044] There are different warehousing processes for structured and unstructured data;

[0045] For structured data, after passing the quality inspection, the data is directly classified and stored according to the constructed railway green performance data knowledge graph;

[0046] For unstructured data, based on the above-mentioned existing data processing processes and image processing processes, the format and content of unstructured data are basically sorted and classified, and then the data is bound to specific subject nodes one by one according to the data classification of the knowledge graph to complete the storage of unstructured data;

[0047] Generate a railway green performance database based on knowledge graph.

[0048] According to a railway green performance data knowledge extraction method provided by the present invention, the data standardization storage process further includes:

[0049] The update of structured stored data is completed by adopting metadata-based data model construction and management technology and dynamic report design method based on extension mechanism.

[0050] The beneficial effects of the railway green performance data knowledge extraction method provided by the present invention are:

[0051] 1. Improve the environmental friendliness of railway projects: By making full use of the "four savings and one environmental protection" data, the performance of railway projects in energy saving, water saving, land saving, material saving and environmental protection can be better monitored and improved;

[0052] 2. Optimize information management: Provide a standardized and systematic knowledge extraction method for the railway industry, so that data collected from different stages and sources can be effectively integrated and used;

[0053] 3. Increase the data-driven power of decision support: Use accurate and refined knowledge to help decision makers understand the actual situation and provide detailed data support as the basis for decision-making, so as to formulate more scientific and reasonable strategies;

[0054] 4. Promote data sharing and communication: By creating a unified knowledge graph platform, stakeholders can more easily access, share and understand relevant data, improving the efficiency of cross-domain and departmental cooperation;

[0055] 5. Improve the sustainability of railway projects: Through continuous tracking and improvement of environmental performance data, the sustainable development capacity of railway projects can be enhanced and the overall impact on the environment can be effectively controlled;

[0056] 6. Achieve compliance with environmental regulations and policies: Accurate data knowledge extraction and management can help railway companies ensure that their projects comply with relevant environmental regulations and standards;

[0057] 7. Improve the timeliness and dynamic update capability of data: With the update of data sources, the automated data integration and update mechanism can ensure the timeliness and accuracy of information, so that railway green data management can quickly respond to changes in the external environment.

[0058] The method of the present invention not only improves the efficiency and accuracy of knowledge extraction, but also makes the monitoring and improvement of key performance indicators of railway engineering more scientific and systematic, further promoting energy conservation and consumption reduction and ecological protection of railway engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0060] Figure 1 It is a flow chart of a railway green performance data knowledge extraction method provided by the present invention;

[0061] Figure 2 It is a schematic diagram of ontology construction in a railway green performance data knowledge extraction method provided by the present invention;

[0062] Figure 3 It is a schematic diagram of the contact network knowledge graph in the railway green performance data knowledge extraction method provided by the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] In one embodiment, see Figure 1-3 , a railway green performance data knowledge extraction method, comprising:

[0065] Step 1: Data collection and preprocessing: collect key data of “four conservations and one environmental protection” in the railway design, construction and operation stages, improve the quality of the collected key data, and conduct preliminary classification;

[0066] Step 2: Knowledge graph construction. The ontology of the entire railway life cycle is constructed mainly using a top-down approach. Knowledge extraction operations are performed according to different data types. The relationship between the extraction results and the ontology is established and stored using a graph database. The repeated and conflicting information in heterogeneous data sources is integrated to generate a knowledge graph.

[0067] Step three, data is standardized and stored in the database. The structured data is stored in the database based on the data model construction and management technology based on metadata. For unstructured data, the format and content of the unstructured data are sorted and classified based on various existing data processing algorithms and image processing algorithms, and then the data is bound to the specific subject nodes one by one according to the data classification of the knowledge graph to complete the storage of the unstructured data.

[0068] The principle of the above embodiment is that there are many sources of basic data on railway green performance and they do not exist independently, and it is necessary to timely organize and store massive structured and unstructured data. The processing of basic data on railway green performance mainly includes the collection of basic data on four savings and one environmental protection, namely energy saving, water saving, land saving, material saving and environmental protection, and the extraction and sorting of structured and unstructured data in the design, construction and operation periods in the form of projects, lines, professions and design units, and the import of data into the database; this embodiment uses knowledge graph (KG) to participate in the knowledge extraction process; KG was first proposed by Google in 2012. After nearly 10 years of development, it has become one of the research hotspots in the field of artificial intelligence. With the help of the naturally adaptive semantic web technology, the knowledge storage based on KG can reflect the way humans acquire and recognize knowledge on the one hand, and on the other hand, it can make it easier for some intelligent systems to solve complex problems. At the same time, thanks to the structured knowledge representation method, KG can facilitate human analysis and reasoning operations on knowledge. In addition, the development of knowledge visualization technology in recent years has given people a more intuitive understanding of knowledge graphs, which undoubtedly provides a basis for people to conduct more in-depth research and application of knowledge graphs, and to a certain extent, it has also promoted knowledge graphs from researchers to the general public; the construction and function introduction of knowledge graphs are shown in Table 1 below:

[0069]

[0070]

[0071] Compared with traditional data processing methods, the advantages and disadvantages of knowledge graphs are shown in Table 2:

[0072]

[0073]

[0074] There are various knowledge graph platforms (KGP) on the market, and the functions of different KGPs vary greatly. For customers, they are less concerned about the internal organizational form of KGP, but more concerned about whether KGP has functions that meet their needs and whether the efficiency of function execution can meet their expectations. Therefore, whether a KGP can be widely used or occupy a larger application market depends to a large extent on whether it can provide sufficient functions and whether the efficiency of the functions is high enough during operation. To a certain extent, it can be said that the KGP function is the part that most directly reflects the application value of KGP. Therefore, when designing and developing KGP, it is necessary to clarify customer needs and make detailed analysis and research on the functions to be developed so as to achieve effective development and efficient use. In order to enable developers to have a deeper understanding of the functional categories of KGP and the relationship between functions. Table 3 introduces the currently commonly used knowledge graph construction platforms:

[0075]

[0076]

[0077] The beneficial effects of the above embodiments are:

[0078] Data integration: Step 1 promotes the integration and unified processing of data in railway design, construction, and operation stages by collecting and preprocessing data, improving data quality and classifying it, and laying a good foundation for subsequent knowledge extraction.

[0079] Knowledge graph construction: Step 2 uses a top-down approach to construct the ontology of the entire life cycle of the railway, and through the combination of knowledge extraction and graph database, it promotes the clarification of the relationship between data and the visualization of knowledge, which helps the connections and deep associations between data to be discovered and utilized.

[0080] Relevance and queryability: By building a knowledge graph and integrating duplicate and conflicting information in heterogeneous data sources, the relevance and queryability of data are enhanced, making analysis and decision support more accurate and efficient.

[0081] Step 3: In the later stage of building the railway green performance data knowledge graph, the railway life cycle ontology is automatically updated with data-driven;

[0082] Data standardization and warehousing: The data standardization and warehousing operation in step 4 ensures that structured data and unstructured data are properly managed and stored, improves the efficiency of data retrieval and use, and also facilitates long-term management and maintenance of data.

[0083] Decision support and optimization: The entire knowledge extraction method helps to provide more accurate data support, provide a scientific basis for green performance evaluation and decision-making in the railway industry, and help comprehensively improve the economic, environmental and social benefits of railway projects.

[0084] Real-time and dynamic: Since data and knowledge are effectively managed, real-time and dynamic data monitoring and analysis can be achieved, providing instant feedback and rapid adjustment capabilities for monitoring and optimizing railway green performance;

[0085] In summary, the extraction method provided in this embodiment enhances the data intelligent processing and knowledge extraction in the field of railway green performance by improving the ability of data integration and analysis, and provides strong support for research and practice in the field of railway environment.

[0086] To further optimize the above embodiment, please refer to Figure 1 ,The data collected in the data collection process are the key data of “Four Savings and One Environmental Protection” in railway engineering, specifically including energy saving data, water saving data, land saving data, material saving data and environmental protection data;

[0087] The above five types of data are collected once and regularly during the design, construction and operation stages of railway projects.

[0088] It should be noted that, taking energy-saving data as an example, its design data source is the design energy-saving report compiled by the railway design unit, which reflects the expected energy consumption overview, energy-saving measures and energy-saving effects during the railway operation period. Since they are relatively concentrated, it is tentatively planned to collect data only through one collection. The specific data is divided into two parts: energy consumption and energy saving. Construction data mainly comes from construction documents and construction sites of typical railway work sites. Operation data mainly comes from existing railway operation data statistical systems, such as statistical indicators of railway energy-saving and environmental protection statistical information systems and statistical data of railway bureau companies along the line. After determining the specific source of the data, it is necessary to arrange relevant personnel to connect the data with the target unit or to collect and calculate relevant data at the project site. In order to further optimize the above embodiment, please refer to Figure 1 , the data preprocessing process includes:

[0089] Format the collected data and convert the files into a processable format; determine the data integrity and quality, and clean irrelevant or erroneous data;

[0090] Standardize the naming of data, clarify the names of specific data contact persons, and make preliminary classifications of data in the form of videos and pictures;

[0091] The preliminary classification process divides the data into structured data and unstructured data according to its structural form; the unstructured data also includes semi-structured data and electronic data. The main way to obtain unstructured data is through surveys conducted by construction units, design units, construction units, supervision units and railway bureau companies.

[0092] It should be noted that after collecting all kinds of data at all stages, it is necessary to pre-process the diversified data, including but not limited to the standardized naming of data files, regularization of data formats, identification of specific data connection personnel, digitization of paper materials, and preliminary classification of video and picture materials; data is divided into structured data, unstructured data, and semi-structured data in between according to its structure. Structured data usually refers to data that can be stored, searched, and analyzed in a relational database, such as text in a table.

[0093] Unstructured data refers to data in the form of videos, images, and sounds, which usually require specific processing and analysis techniques to understand its content.

[0094] Semi-structured data refers to data in an intermediate state, such as e-mails and XML documents, which contain both structured elements and a large amount of unstructured text.

[0095] The preprocessing stage needs to be closely coordinated with the subsequent knowledge graph construction and data standardization steps. The preprocessed data will directly affect the quality of the knowledge graph and the application effect of the final data. A good preprocessing process can greatly reduce the burden of subsequent steps and improve the accuracy and efficiency of knowledge extraction.

[0096] To further optimize the above embodiment, please refer to Figure 1-3 , the knowledge graph construction process includes:

[0097] The ontology construction mainly adopts the top-down method, using the data of railway design, construction and operation stages and the suggestions of experts in this field to establish the railway life cycle ontology;

[0098] Knowledge extraction: quality inspection of structured data and extraction based on rules; named entity recognition, relationship extraction, and attribute extraction are required for unstructured data to generate knowledge content for each data;

[0099] Knowledge fusion: store the knowledge content in the knowledge data framework, and perform data fusion when there is data duplication and conflict; associate the knowledge data framework with the railway life cycle ontology, improve and generate the entity relationship network, and perform quality assessment based on the current updated knowledge;

[0100] Generate a graph. When the quality assessment results meet the preset conditions, a railway green performance data knowledge graph is generated based on the entity relationship network.

[0101] It should be noted that ontology originally originated from philosophy. It is a description of the essence of objective things and is used to describe and classify objective and real things. With the development of technologies such as knowledge graph, ontology has been increasingly valued by researchers; from describing the objective world with the help of ontology, ontology has shifted to summarizing entities and relationships in the knowledge graph field; by describing the concepts and relationships between concepts in a general field or a specific field, the description of this field is completed with knowledge graph; to establish a knowledge graph, the knowledge in the railway field must first be abstracted and organized to obtain the ontology. The purpose of constructing the ontology is to standardize different entities, attributes and relationships; ontology is the core issue of constructing the knowledge graph, and the construction of the ontology is also guided by the problems that the knowledge graph is to solve;

[0102] There are three main types of ontology construction: top-down method, bottom-up method and comprehensive method. This technical solution is to establish the ontology in the field of railway engineering. The entities in this field are highly specialized and require high accuracy, so the top-down method is adopted to construct the ontology. By referring to the relevant information of the railway design, construction and operation stages, and then combining the suggestions of railway experts, a schematic diagram of the railway life cycle ontology is established. Figure 2 ;

[0103] The knowledge graph proposed in this embodiment is aimed at the collection, processing and storage of green performance data of the entire life cycle of railways, covering the three aspects of railway design, construction and operation. Therefore, the construction of the ontology should be comprehensive and face the specific business object of the integrated application of railway green performance related data;

[0104] In the process of knowledge extraction, relevant algorithms are used to extract electronic data;

[0105] Among them, paper materials, video images, and voice materials in electronic data are processed using text recognition, image recognition, and voice recognition algorithms to generate text, and then knowledge is extracted using text processing operations;

[0106] In the process of knowledge extraction, the extraction process for unstructured data includes:

[0107] Named entity recognition, where a named entity refers to an entity with a specific meaning in the text. The recognition process is to identify three major categories and seven subcategories in the text;

[0108] Relation extraction, determining the semantic relationship between two or more entities after identifying the named entities;

[0109] Attribute extraction, used to obtain specific attributes of an entity;

[0110] Knowledge integration, combining the identified named entities, relations, and attributes to form knowledge content;

[0111] The goal of knowledge extraction is to extract knowledge from massive amounts of data through technical means. The methods of knowledge extraction vary according to the type of knowledge data to be extracted. Knowledge extraction mainly includes named entity recognition, relationship extraction, and attribute extraction. Different extraction methods are used for data in different formats. Structured data can be extracted based on rules. For unstructured data, named entity recognition, relationship extraction, and attribute extraction are required to obtain structured knowledge. For electronic data, relevant algorithms can be used directly to extract knowledge. For a large amount of paper, image, voice and other materials generated during the entire life cycle of the railway, image recognition technology can be used to apply text recognition, image recognition, and voice recognition technology to paper materials, video images, voice and other materials to process them into text, and then use text processing operations to extract knowledge.

[0112] In named entity recognition, named entities refer to entities with specific meanings in texts, which generally refers to identifying three major categories and seven minor categories in texts. For example, using OCR technology to identify and convert railway engineering related image data into text, such as "the amount of earth excavated is XXX cubic meters", the image is analyzed and classified according to the text content to determine the specific name and entity type of the image data.

[0113] The knowledge fusion process includes:

[0114] After the knowledge content is extracted, a graph database is used to store the knowledge content in a knowledge data framework, which is a graph of points and edges for users and a low-dimensional dense vector for computers;

[0115] For data of different modalities and sources, there will be certain conflicts and duplications. It is necessary to integrate knowledge according to different information sources, and complete and remove duplicates from the knowledge graph. Among them, knowledge integration methods include ontology alignment and entity alignment.

[0116] Ontology alignment is used to determine whether elements or structures in two knowledge data frameworks should match;

[0117] Entity alignment is used to map entities in knowledge content into the same vector space and to achieve entity disambiguation by calculating entity distances;

[0118] After the ontology alignment and entity alignment process is completed, knowledge reasoning is performed on the alignment results, and the alignment results and reasoning results are associated with the railway life cycle ontology; an entity relationship network is generated and quality assessment is performed; if there is newly extracted knowledge content, the entity relationship network is updated and the quality assessment is repeated;

[0119] After knowledge is extracted, the entities and relationships are stored in a graph database, which is a graph of points and edges for users and a low-dimensional dense vector for computers. There will be certain conflicts and duplications for data of different modalities and sources. It is necessary to fuse knowledge according to different information sources, and complete and remove duplicates from the knowledge graph. Common knowledge fusion methods include ontology alignment and entity alignment.

[0120] Ontology alignment is used to determine whether elements in two knowledge graphs should match, without considering the matching of other elements; ontology alignment is to solve the problem of joint use of different knowledge graphs. In the process of building the knowledge graph of green performance data of railway engineering, the existing general knowledge graph resources can be effectively utilized; according to different practical technologies, ontology alignment can be divided into two levels: element level and structure level;

[0121] Entity alignment means that for the same entity, there will be different expressions in different data sources. In order to eliminate the ambiguity caused by different naming in the knowledge graph, all entities are mapped to the same vector space, and entity disambiguation is achieved by calculating the entity distance. The specific method is to find a corresponding set with equivalent relations in this set given two entity sets, such as the contact network from site A to site B can be defined as section 001 contact network;

[0122] Furthermore, a knowledge graph can be obtained; taking the main contact network of railway power support as an example, a railway green performance knowledge graph is constructed, such as Figure 3 As shown in the figure, as the power supply system of electrified railways, the overhead contact network mainly includes two aspects of data, namely, consumable data and energy consumption data. The consumable data specifically refers to the data of various parts contained in the overhead contact network, and the energy consumption data refers to the power consumption data of the overhead contact network.

[0123] To further optimize the above embodiment, please refer to Figure 1,In the later stage of the construction of the railway green performance data knowledge graph, the railway life cycle ontology can be updated automatically with data-driven. Along with the updates of equipment, personnel, and data subjects, new ontology classifications will appear, and the ontology will be expanded through data-driven automatic updates;

[0124] The data-driven automatic update of the railway life cycle ontology includes the following processes:

[0125] Continuously monitor incoming data streams from various sensors, devices, and systems, automatically identifying and capturing new instances or changes;

[0126] Use machine learning and natural language processing algorithms to automatically extract new knowledge from real-time monitored data for expansion or modification of existing ontologies;

[0127] On top of automated updates, an expert review process is set up, where experts can manually review and adjust the ontology updates recommended by the algorithm.

[0128] Set up a data conflict mechanism to integrate or adjust the old ontology classification based on the access of new data sources;

[0129] Set up a relationship management mechanism to manage the relationship between historical data and new data to ensure a smooth transition between old and new data;

[0130] Set up reasoning mechanisms to infer new associations and insights from existing information.

[0131] It should be noted that in the later stage of the construction of the railway green performance data knowledge graph, data-driven automatic updates are crucial to maintaining the timeliness and accuracy of the graph. Data-driven automatic updates of the railway life cycle ontology include the following aspects:

[0132] Real-time monitoring continuously monitors incoming data streams to identify new instances or changes, such as new equipment, personnel transfers, or changes in data subjects.

[0133] Automated ontology learning uses algorithms to automatically extract new knowledge from data and expand or modify the ontology. For example, new railway equipment may bring new performance parameters, and personnel skill upgrades may involve new responsibilities.

[0134] Semi-automatic ontology maintenance, although updates can be automated, a mechanism needs to be left for experts to manually review and modify automatically recommended ontology updates to ensure their quality and accuracy.

[0135] Data fusion and accuracy maintenance, with the access of new data sources, the old ontology classification may need to be integrated or adjusted to ensure data consistency. Therefore, the automated update system also needs to be equipped with a mechanism to resolve data conflicts and ensure accuracy.

[0136] Evolution management,The expansion of the ontology should consider not only adding new categories or attributes, but also how to manage the relationship between historical data and new data.,A smooth transition between old and new data is necessary to maintain the integrity and usefulness of historical records.

[0137] Intelligent reasoning ability: In order to adapt to the new ontology classification and ensure the effective use of the knowledge graph, the updated knowledge graph should have intelligent reasoning ability and be able to infer new associations and insights from existing information;

[0138] In some embodiments, the specific implementation is as follows:

[0139] 1. Real-time data flow monitoring and instance capture

[0140] Various sensors (such as temperature, humidity, vibration sensors, etc.) are installed at key locations along the railway, and an equipment status monitoring system is integrated to collect real-time environmental, equipment health and operation data.

[0141] Build an efficient data stream processing platform, such as using Apache Kafka or Amazon Kinesis, to collect, store, and initially process real-time data to ensure the timeliness and availability of data.

[0142] 2. Automated ontology learning and expansion

[0143] Use deep learning models (such as convolutional neural networks (CNN) or recurrent neural networks (RNN)) and natural language processing techniques (such as BERT) to automatically extract patterns and knowledge from real-time data and identify new entity types or attributes.

[0144] The algorithm is designed to compare the extracted new knowledge with the existing ontology, and adopt semantic similarity algorithms (Word2Vec / Doc2Vec, Transformer model, Ontology Alignment tool, etc.) to identify potential classification extension points and realize automatic expansion of the ontology.

[0145] 3. Semi-automatic ontology maintenance and expert review

[0146] A user-friendly interface is set up to allow domain experts to review ontology update suggestions made by the algorithm, including new categories, attributes, or relations.

[0147] Experts can suggest modifications or make adjustments directly, and the system records this feedback to optimize subsequent automated learning algorithms.

[0148] 4. Data fusion and accuracy maintenance

[0149] Design algorithms to resolve data conflicts, such as using statistical methods or expert rules, to determine the most appropriate classification or attribute value.

[0150] Implement data cleansing and validation processes to remove duplicates, errors, or inconsistencies to ensure data accuracy and consistency.

[0151] 5. Evolutionary management and relationship maintenance

[0152] Develop conversion rules, including attribute mapping tables (create attribute mapping tables to clearly indicate the correspondence between old attributes and attributes in the new ontology), data migration scripts (write SQL or Python scripts to perform data conversion to ensure that historical data is correctly migrated according to the new classification), and data backtracking mechanisms (add data backtracking logic to the conversion rules so that the original meaning of historical data can still be accessed and understood under the new ontology structure); map historical data to the new ontology classification to ensure seamless connection between historical data and new data.

[0153] Establish a data relationship management mechanism, including a version control system (similar to Git, used to track every change to data and ontology structure for easy auditing and recovery), a relationship graph database (using a graph database such as Neo4j or JanusGraph to store and query complex relationships between entities to ensure connectivity between data), and an event log (recording the timestamp, change content, and operator of each data change to facilitate future analysis and auditing); track data change history to ensure that the relationship between new and old data is clear for easy tracing and analysis.

[0154] 6. Intelligent Reasoning and Insight

[0155] Introduce knowledge graph reasoning engines, such as OWL2-RL or SWRL, to automatically deduce implicit relationships and rules between data.

[0156] Combining expert knowledge and data analysis, valuable insights are generated from the updated knowledge graph, such as predicting equipment failure trends or optimizing operation and maintenance strategies.

[0157] Through the above implementation methods, a flexible, intelligent and continuously evolving railway life cycle ontology can be built to effectively support the management and analysis of green performance data and improve the sustainable development level of the railway system.

[0158] The algorithms in the above embodiments are prior art and their principles are not elaborated here.

[0159] In order to further optimize the above embodiment, the data standardization storage process includes:

[0160] There are different warehousing processes for structured and unstructured data;

[0161] For structured data, after passing the quality inspection, the data will be directly classified and stored according to the constructed railway green performance data knowledge graph;

[0162] For unstructured data, based on the above-mentioned existing data processing processes and image processing processes, the format and content of unstructured data are basically sorted and classified, and then the data is bound to specific subject nodes one by one according to the data classification of the knowledge graph to complete the storage of unstructured data;

[0163] Generate a railway green performance database based on knowledge graph.

[0164] It should be noted that the data standardization warehousing process also includes:

[0165] The update of structured stored data is completed by using metadata-based data model construction and management technology and dynamic report design method based on extension mechanism;

[0166] After collecting, preprocessing, and constructing the knowledge graph of railway engineering green performance data, the data needs to be standardized and stored in the database to facilitate subsequent data query and analysis. There are different storage processes for structured data and unstructured data. The format of structured data is relatively regular. Therefore, after the data is checked for qualification, it can be directly classified and stored according to the constructed railway green performance data knowledge graph. For the update of structured stored data, this technical solution mainly adopts: (1) metadata-based data model construction and management technology; (2) dynamic report design method based on extension mechanism;

[0167] Among them, (1) the principles of data model construction and management technology based on metadata:

[0168] Metadata is data about data, which is used to describe the attributes, sources, usage and control information of data. Metadata-based data model construction and management technology refers to the use of metadata to define, build and manage data models. In this way, metadata not only describes the structure of data, but also contains rules on how to process and operate data. The key advantages of this technology are:

[0169] Dynamicity: The system can dynamically build data models based on metadata descriptions to adapt to rapidly changing business needs;

[0170] Flexibility: Developers can adjust the data model by modifying metadata instead of code, which improves the system's adaptability and responsiveness;

[0171] Maintainability: System maintainers can more easily understand and manage data models because metadata provides rich contextual information;

[0172] Metadata-based management platforms usually allow users to define entity types, relationships, attributes, and related business rules;

[0173] (2) The principle of dynamic report design method based on extension mechanism:

[0174] In information systems, reports are an important means of displaying and analyzing data. Traditional report design is usually static, that is, the format and content of the report are fixed at the beginning of the design. However, in actual business, users' needs for reports are often changeable. In order to cope with such changes, a dynamic report design method based on an extension mechanism has emerged.

[0175] This method allows users to dynamically create or adjust reports by adding, modifying, or deleting report elements (such as fields, charts, formats, etc.) on the report template according to their needs; this design method has the following characteristics:

[0176] User self-service: Users can customize reports as needed without the intervention of developers;

[0177] Immediacy: Report changes can be generated instantly to quickly respond to business needs;

[0178] Flexibility: can cope with complex and changing reporting requirements;

[0179] Dynamic report design usually relies on powerful report engines and interactive report design tools, which provide graphical interfaces for users to design reports through simple drag-and-drop and configuration operations; Wheel Assistant explains here, as well as using advanced features such as data binding, parameter passing, formula calculation, etc. to enhance the functionality of reports;

[0180] The above embodiments greatly improve the flexibility of databases and information systems and user experience, allowing users with non-technical backgrounds to quickly build and modify data models and reports according to their needs;

[0181] Furthermore, for the storage of unstructured data, the format and content of structured data are basically sorted and classified based on various existing data processing algorithms and image processing algorithms, and then the data are bound to specific subject nodes one by one according to the data classification of the knowledge graph to complete the storage of unstructured data.

[0182] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for extracting knowledge from railway green performance data, characterized in that: include: Step 1: Data collection and preprocessing: collect key data of "four conservations and one environmental protection" in the railway design, construction and operation stages, improve the collected key data, and conduct preliminary classification; Step 2: Knowledge graph construction. The ontology of the entire railway life cycle is constructed mainly using a top-down approach. Knowledge extraction operations are performed according to different data types. The relationship between the extraction results and the ontology is established and stored using a graph database. The repeated and conflicting information in heterogeneous data sources is integrated to generate a knowledge graph. Step 3: In the later stage of building the railway green performance data knowledge graph, the railway life cycle ontology is automatically updated with data-driven; Step 4: Standardize and store data in the database, and use metadata-based data model construction and management technology to store structured data in the database; For unstructured data, based on various existing data processing algorithms and image processing algorithms, the format and content of unstructured data are sorted and classified, and then the data is bound to specific subject nodes one by one according to the data classification of the knowledge graph to complete the warehousing of unstructured data.

2. A railway green performance data knowledge extraction method according to claim 1, characterized in that: The data collected during the data collection process are the key data of "four savings and one environmental protection" in railway engineering, specifically including energy-saving data, water-saving data, land-saving data, material-saving data and environmental protection data; The above five types of data are collected once and regularly during the design, construction and operation stages of railway projects.

3. A railway green performance data knowledge extraction method according to claim 2, characterized in that: The data preprocessing process includes: Format the collected data and convert the files into a processable format; determine the data integrity and quality, and clean irrelevant or erroneous data; Standardize the naming of data, clarify the names of specific data contact persons, and make preliminary classifications of data in the form of videos and pictures; The preliminary classification process divides the data into structured data and unstructured data according to its structural form; the unstructured data also includes semi-structured data and electronic data. The main way to obtain unstructured data is through surveys conducted by construction units, design units, construction units, supervision units and railway bureau companies.

4. A railway green performance data knowledge extraction method according to claim 3, characterized in that: The knowledge graph construction process includes: The ontology construction mainly adopts the top-down method, using the data of railway design, construction and operation stages and the suggestions of experts in this field to establish the railway life cycle ontology; Knowledge extraction: quality inspection of structured data and extraction based on rules; named entity recognition, relationship extraction, and attribute extraction are required for unstructured data to generate knowledge content for each data; Knowledge fusion: store the knowledge content in the knowledge data framework, and perform data fusion when there is data duplication and conflict; associate the knowledge data framework with the railway life cycle ontology, improve and generate the entity relationship network, and perform quality assessment based on the current updated knowledge; Generate a graph, and when the quality assessment result meets the preset conditions, generate a railway green performance data knowledge graph based on the entity relationship network.

5. A railway green performance data knowledge extraction method according to claim 4, characterized in that: In the later stage of the construction of the railway green performance data knowledge graph, the railway life cycle ontology is automatically updated with data-driven. Along with the updates of equipment, personnel, and data subjects, new ontology classifications will appear, and the ontology will be expanded through data-driven automatic updates. The data-driven automatic update of the railway life cycle ontology includes the following processes: Continuously monitor incoming data streams from various sensors, devices, and systems, automatically identifying and capturing new instances or changes; Use machine learning and natural language processing algorithms to automatically extract new knowledge from real-time monitored data for expansion or modification of existing ontologies; On top of automated updates, an expert review process is set up, where experts can manually review and adjust the ontology updates recommended by the algorithm. Set up a data conflict mechanism to integrate or adjust the old ontology classification based on the access of new data sources; Set up a relationship management mechanism to manage the relationship between historical data and new data to ensure a smooth transition between old and new data; Set up reasoning mechanisms to infer new associations and insights from existing information.

6. A railway green performance data knowledge extraction method according to claim 5, characterized in that: In the process of knowledge extraction, the electronic data is extracted using relevant algorithms; Among them, paper materials, video images, and voice materials in electronic data are processed using text recognition, image recognition, and voice recognition algorithms to generate text, and then knowledge is extracted using text processing operations.

7. A railway green performance data knowledge extraction method according to claim 6, characterized in that: In the knowledge extraction process, the extraction process for unstructured data includes: Named entity recognition, where a named entity refers to an entity with a specific meaning in the text. The recognition process is to identify three major categories and seven subcategories in the text; Relation extraction, determining the semantic relationship between two or more entities after identifying the named entities; Attribute extraction, used to obtain specific attributes of an entity; Knowledge integration combines the identified named entities, relations, and attributes to form knowledge content.

8. A railway green performance data knowledge extraction method according to claim 7, characterized in that: The knowledge fusion process includes: After the knowledge content is extracted, the knowledge content is stored in the knowledge data framework using a graph database, which is a graph of points and edges for users and a low-dimensional dense vector for computers; For data of different modalities and sources, there will be certain conflicts and duplications. It is necessary to integrate knowledge according to different information sources, and complete and remove duplicates from the knowledge graph. Among them, knowledge integration methods include ontology alignment and entity alignment. The ontology alignment is used to determine whether elements or structures in two knowledge data frameworks should match; The entity alignment is used to map all entities in the knowledge content into the same vector space, and to achieve entity disambiguation by calculating the entity distance; After the ontology alignment and the entity alignment process are completed, knowledge reasoning is performed on the alignment results, and the alignment results and reasoning results are associated with the railway life cycle ontology; the entity relationship network is generated and quality assessment is performed; if there is newly extracted knowledge content, the entity relationship network is updated and the quality assessment is repeated.

9. A railway green performance data knowledge extraction method according to claim 8, characterized in that: The data standardization warehousing process includes: There are different warehousing processes for structured and unstructured data; For structured data, after passing the quality inspection, the data is directly classified and stored according to the constructed railway green performance data knowledge graph; For unstructured data, based on the above-mentioned existing data processing processes and image processing processes, the format and content of unstructured data are basically sorted and classified, and then the data is bound to specific subject nodes one by one according to the data classification of the knowledge graph to complete the storage of unstructured data; Generate a railway green performance database based on knowledge graph.

10. A railway green performance data knowledge extraction method according to claim 9, characterized in that: The data standardization warehousing process also includes: The update of structured stored data is completed by adopting metadata-based data model construction and management technology and dynamic report design method based on extension mechanism.

Images