Intelligent diagnosis-oriented device resource information ontologized sharing platform

Abstract

The application provides an equipment resource information ontologization sharing platform for intelligent diagnosis, comprising a service layer and a presentation layer, in which data analysis, ontology construction, rule reasoning and data management are completed; the rule reasoning is knowledge reasoning for equipment quality information resource ontology, comprising: data feature screening is performed by using a feature screening model, the feature screening model is an Apriori model based on distributed frequent item set mining and association rule promotion degree screening, modeling of the equipment quality information resource ontology is performed based on screened data features, and through construction of feature mapping, distributed matching, double-condition rule triggering and weighted fusion and confidence driven updating mechanism, dynamic reasoning and adaptive updating of the ontology model knowledge associated with the equipment quality features are performed; the presentation layer comprises a fault diagnosis module for fault diagnosis and risk prediction. Through ontology modeling of the equipment use quality information, the application promotes fusion and interaction of information.

Description

Technical Field

[0001] This application relates to the field of equipment resource information management technology, and more particularly to an equipment resource information ontology sharing platform for intelligent diagnostics. Background Technology

[0002] In the current technological environment, equipment resource information ontology sharing platforms mainly rely on network communication technology, database management systems, and cloud computing to achieve resource sharing and data interaction. They construct a unified network topology optimization information display, but they have significant shortcomings in quality control and dynamic knowledge iteration in intelligent diagnostic scenarios. Despite the progress made by existing platforms in resource sharing and data interaction, core issues still need to be addressed: First, there is a lack of structured knowledge for quality diagnosis: equipment quality data (fault characteristics, maintenance records, operating parameters, etc.) lacks unified ontology modeling, and the semantic descriptions of different manufacturers and scenarios are heterogeneous (e.g., the classification rules for "insulation aging" differ in different standards), making it difficult to accurately mine the correlation of quality features and failing to provide a semantically interoperable knowledge foundation for intelligent diagnosis. Second, there is inefficiency in dynamic diagnostic reasoning: existing platforms rely on static rules and do not combine ontology semantic constraints with distributed computing optimization, resulting in feature redundancy, false association interference, and an inability to dynamically adjust diagnostic logic based on the statistical characteristics of real-time equipment operation, leading to lagging reasoning conclusions. Finally, there is a lack of a closed-loop knowledge update: diagnostic conclusions cannot effectively feed back into ontology knowledge iteration, and the ontology model is difficult to adaptively update in the face of changes in the equipment's state throughout its entire lifecycle (e.g., aging, operating condition fluctuations), leading to a continuous decline in diagnostic accuracy. Summary of the Invention

[0003] The purpose of this application is to provide a platform for sharing equipment resource information ontology for intelligent diagnostics, thereby addressing at least the technical problems of the lack of structured quality diagnostic knowledge, the difficulty in adaptively updating ontology models, and the decline in diagnostic accuracy in related technology sharing platforms. To achieve the above objective, the present invention provides the following technical solution.

[0004] According to one embodiment of this application, a device resource information ontology sharing platform for intelligent diagnostics is provided, including: The business layer receives task requests from the presentation layer, decomposes these requests into business requirements, and sends these requirements to the service layer. The service layer then performs data querying and storage based on these requirements, completing data analysis, ontology construction, rule reasoning, and data management. Rule reasoning involves knowledge reasoning for the equipment quality information resource ontology, including: using a feature filtering model to filter data features (the Apriori model based on distributed frequent itemset mining and association rule lifting); modeling the equipment quality information resource ontology based on the filtered data features; and dynamically reasoning and adaptively updating the ontology model knowledge associated with the equipment quality features through feature mapping, distributed matching, dual-condition rule triggering, weighted fusion, and confidence-driven update mechanisms. The presentation layer is used to provide users with an information interaction entry point, receive and display processing results from the business layer, and send task requests to the business layer. The presentation layer includes a user management module, a human-computer interaction module, an information display module, a fault diagnosis module, and an inference rule module. In the fault diagnosis module of the presentation layer, fault diagnosis and risk prediction are performed based on the statistical characteristics of the rule base and fault information. The final risk level is calculated by using statistical values ​​and loss values. Specifically, the current date is obtained, and the difference between the current date and the last maintenance time for each type of equipment is calculated. The percentile of the query result for that type of equipment is used as the statistical value for that type of equipment. The loss value is the percentile of the difference between the current date and the equipment issuance date for all equipment.

[0005] Preferably, the step of data feature selection based on the improved Apriori algorithm includes: From the equipment quality information resource ontology, equipment fault data containing features and fault tags are extracted, and combined with historical data from the fault diagnosis module of the presentation layer, the equipment fault data is organized into transaction form to obtain equipment fault transaction set. Each transaction contains equipment feature items, fault tags and auxiliary diagnostic parameters. Generating frequent itemsets layer by layer: Determine a minimum support threshold to filter frequent itemsets, where support represents the frequency of a feature itemset in all transactions; starting with a single feature item, count the number of times the feature item appears in all transactions in a distributed parallel manner, calculate the feature item support, and retain feature items with support greater than or equal to the minimum support threshold as frequent itemsets; repeat this process until no new frequent itemsets can be generated. Association rule mining and filtering: Determine a minimum confidence threshold, which is used to filter valid association rules. The confidence level represents the probability that a transaction containing feature itemset X also contains feature itemset Y. For each frequent itemset, generate all possible association rules and calculate the confidence level of the frequent itemset. Retain association rules with a confidence level greater than or equal to the minimum confidence threshold. Generate all possible association rules and filter them. Based on the frequently mined itemsets and association rules, data features that have a significant impact on equipment fault reasoning and diagnosis are identified.

[0006] Preferably, in the step of generating frequent itemsets layer by layer: The set of equipment failure transactions is horizontally divided into n data subsets {D1, D2, ..., Dn} according to the service layer load balancing strategy. The n data subsets are then distributed to m computing nodes, where m ≤ n, to build a distributed parallel mining environment. Each computing node executes the Map function on its assigned local subset Di; this involves scanning Di to extract feature terms and counting the occurrences of feature term t in Di. Output intermediate key-value pairs: Key=t, Value= Combined with local minimum support threshold Calculate local support: In the formula, This represents the total number of transactions in the local subset Di; Each computing node executes the Combiner function to aggregate the occurrence counts of the same feature term t. The total number of occurrences of feature t in Di is obtained. Filter to meet The feature terms are used to generate a local 1-frequent itemset for node i. , ; Use the Reduce function to aggregate all nodes The number of times the statistical characteristic term t appears in the total number of transactions. ,in, In the formula, m is the total number of computation nodes; combined with the global minimum support threshold. Calculate global support ,in, Represents the total number of all transactions; filters that meet the criteria. Feature terms, generate global k-frequent itemsets , ; Global (k-1) frequent itemsets The itemset is distributed to each computing node, and each node performs a self-join operation on its local subset Di to generate k-itemset candidates. Combined with ontology association rule constraints, filtering Unreasonable itemsets are removed using Apriori properties. Itemsets containing infrequent (k-1) subsets; local support of candidate sets after pruning is calculated by traversing Di at each node. Global support is calculated through Reduce aggregation. ,reserve Itemsets, generate global k-frequent itemsets ; When the iteration meets the preset conditions, the iteration stops and the global frequent itemset set is output. .

[0007] Preferably, in the association rule mining and filtering step, a lift metric is introduced for filtering association rules, where the lift metric is expressed as: ;in, This represents the lifting degree of feature itemset X to feature itemset Y. It represents the confidence level of feature itemset X with respect to feature itemset Y, and is used to measure the probability that feature itemset Y will also appear when feature itemset X appears. Indicates the support of the feature set Y; when A value greater than 1 indicates that feature itemset X has a positive correlation with feature itemset Y, and it is preferred to retain it. Association rules with a value greater than 1.

[0008] Preferably, in the step of dynamically reasoning and adaptively updating the ontology model knowledge associated with equipment quality features through a feature mapping, distributed matching, dual-condition rule triggering, weighted fusion, and confidence-driven update mechanism: Based on the equipment quality features selected by the improved Apriori algorithm, the mapping relationship between feature vectors and ontology concepts is constructed through the OWL interface of the Jena inference engine; the real-time operation data of the equipment is matched in parallel with the preset fault feature templates in the ontology using a distributed computing framework to generate preliminary matching results. By combining SWRL rules with statistical characteristic indicators, a two-condition mechanism is constructed. When the percentile of the statistical value of a certain feature in the distributed matching result exceeds the preset threshold, the Drools inference engine is activated to call the corresponding SWRL rule set and output the intermediate inference conclusion. The random forest algorithm is introduced to calculate the confidence weight of different inference paths. The distributed matching results and rule-based inference conclusions are weighted and fused. If the overall confidence of the fused results is ≥0.85, the ontology knowledge is updated directly. If the confidence is between 0.7 and 0.85, the human-computer interaction module is triggered for manual confirmation before updating. If the confidence is <0.7, the initial matching results are generated by rematching. When the number of rematches exceeds a preset number, manual intervention is triggered.

[0009] Preferably, in the step of combining SWRL rules with statistical characteristic indicators to construct a two-condition mechanism: The dual-condition mechanism includes the coordinated triggering of rule conditions and statistical threshold conditions; Among them, the rule conditions are constructed based on the object attributes and data attributes in the equipment quality information resource ontology, and adopt the SWRL rule form. The antecedent includes ontology concept association, and the consequent corresponds to the preset diagnostic conclusion. The statistical threshold condition is a combined threshold of the percentile of the statistical value and the percentile of the loss value. Both the percentile of the statistical value and the percentile of the loss value are divided into three levels: high, medium, and low. The two threshold levels are determined by the distribution characteristics of historical fault data. When the ontology concepts in the rule conditions are completely matched and at least one indicator in the statistical threshold conditions reaches the high level, the dual-condition mechanism is triggered, activating the corresponding SWRL rule set. If the rule conditions are matched but the statistical threshold conditions only meet the medium level, the simplified rule set is triggered. If the statistical threshold conditions are lower than the medium level, the rule is not triggered temporarily, and only the matching result is recorded for subsequent threshold calibration. After the rule set is triggered, the specific values ​​of the statistical threshold conditions are embedded into the reasoning process through the ontology inference engine tool to generate a structured conclusion containing a triple of fault characteristics, statistical indicators, and handling measures. The statistical indicator values ​​are synchronously updated to the corresponding data attribute fields of the equipment quality information resource ontology, realizing the dynamic association between rule reasoning and ontology instances.

[0010] Preferably, the information display module is used to visualize information that matches the query conditions input by the user in the interaction, and to query, add, update and delete the matched information; In the information display module, the information displayed visually includes equipment information, test information, maintenance information, and fault information; Equipment information describes the composition of the equipment, including equipment model, equipment name, equipment number, equipment composition, assembly information, production information, and equipment distribution date. Test information includes the object under test, test items, testers, test results, test instruments, and test time; Maintenance information includes equipment maintenance records and maintenance manual information. Equipment maintenance records include maintenance objects, maintenance methods, maintenance content, and maintenance dates. The fault information is used to describe the historical fault data of the equipment and is divided into two parts: fault category and maintenance category. The fault category includes fault type, fault phenomenon, fault location, fault cause and fault time information; the maintenance category includes maintenance method, maintenance personnel and maintenance time information.

[0011] Preferably, in the service layer, a complete ontology includes classes, object relationships, data relationships, and instances; The ontology construction process includes the following steps: In a single-structure scenario, a root class is created, corresponding data attributes are defined for each class, and object attributes are created to represent the relationships between classes. In the case of complex structures, create classes for the complex structures and define data attributes for the complex structures; Create an instance for each class that is created.

[0012] Preferably, in the step of performing knowledge reasoning for the equipment quality information resource ontology using rule-based reasoning, the completed ontology project is exported as an OWL format file. The ontology reasoning process is then implemented based on the OWL format file, including: reading the OWL file of the created equipment quality information resource ontology using the Jena inference engine; registering and creating an inference engine according to the rules of the equipment quality information resource using the API provided by the Jena inference engine; binding the equipment quality information resource ontology and the inference engine together to generate a model object; calling the Ontology API and Model API through the model object to perform reasoning operations on the equipment quality information resource ontology and obtaining the reasoning results; obtaining the reasoning results through the API provided by Jena; and updating the equipment quality information resource ontology using the Jena API based on the reasoning results, wherein the update operations include adding new information, modifying existing information, deleting redundant information, and saving the updated ontology.

[0013] Preferably, in the fault diagnosis module, the method for performing fault diagnosis includes the following steps: Establish a risk database, which includes data types, equipment models, equipment names, maintenance times, and failure times. The equipment is classified according to its model and name. Based on the classification results, the historical failure information of the equipment is statistically analyzed. The statistical information includes the failure frequency, maintenance frequency, and maintenance-failure time difference of the equipment. For each type of equipment, query the time of the last maintenance before the failure based on the failure time in each failure data. If there is no maintenance data before the failure, the maintenance time is equal to the failure time, the difference between the two is calculated, and the differences are sorted in ascending order. The sorted results are stored in the database for later use. Calculate the final risk level in response to a user's click that triggers a risk prediction action in the interactive interface.

[0014] Compared with existing technologies, the beneficial effects of this invention's device resource information ontology sharing platform for intelligent diagnostics are: First, this application breaks down device data silos through ontological modeling. By establishing structured associations of classes, attributes, and instances, it achieves semantic interoperability of quality information across devices and throughout the entire lifecycle, providing a unified knowledge foundation for intelligent diagnosis. The improved Apriori algorithm combines distributed computing with two-layer support filtering to accurately mine high-value association rules in large-scale device data, balancing processing efficiency and feature reliability, and solving the pain points of redundant feature interference and insufficient single-machine computing power in traditional algorithms. Second, this application adopts a reasoning architecture of feature mapping-distributed matching-dual-condition rule triggering-confidence weighted fusion, which deeply integrates SWRL semantic rules with statistical characteristic indicators. Through multi-engine collaboration and dynamic threshold mechanism, the diagnostic decision can not only adapt to the real-time status changes of the device, but also achieve flexible updates through automatic-human-machine collaboration through confidence layering, thus breaking through the limitations of traditional static rule reasoning. Third, the presentation layer of this application provides users with an intuitive interface through the human-computer interaction module and information display module, enhancing the user's interactive experience; the inference engine module of the service layer performs semantic reasoning and analysis based on ontology models and rules, enabling users to provide more in-depth analysis of equipment resource information; during the fault reasoning process, the rule set is applied to the knowledge base, and a data model is constructed by analyzing the statistical characteristics of historical fault information, while considering both rule-based reasoning and case-based reasoning schemes to improve the accuracy of reasoning; this invention introduces ontology technology to perform ontology modeling of equipment usage quality information, solving the problem of data heterogeneity, eliminating semantic ambiguity for the same information by different subjects, and promoting information fusion and interaction. Attached Figure Description

[0015] The accompanying drawings, which are provided to further illustrate this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application.

[0016] In the attached diagram: Figure 1 This is a structural block diagram of the device resource information ontology sharing platform for intelligent diagnostics, as described in this application. Figure 2 A schematic diagram illustrating the implementation process of the data feature filtering step using a feature filtering model, provided in an embodiment of this application; Figure 3 This is a schematic diagram illustrating the implementation process of the fault diagnosis method provided in this application embodiment. Detailed Implementation

[0017] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0018] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0019] Please refer to Figure 1 In one embodiment of this application, a device resource information ontology sharing platform for intelligent diagnostics is provided; It includes a presentation layer 11, a business layer 12, a service layer 13, and a data layer 14; among which: In this embodiment of the application, the business layer 12 is used to receive the task request from the presentation layer 11, decompose the task request into business requirements, send the business requirements to the service layer 13, and the service layer 13 performs data query and storage based on the business requirements to complete data analysis, ontology construction, rule reasoning and data management. In this embodiment of the application, rule-based reasoning is used to perform knowledge reasoning on the equipment quality information resource ontology, including: using a feature filtering model to filter data features, wherein the feature filtering model is an Apriori model based on distributed frequent itemset mining and association rule lifting degree filtering, and modeling the equipment quality information resource ontology based on the filtered data features. During the reasoning process, the embodiments of this application dynamically reason and adaptively update the ontology model knowledge associated with equipment quality features by constructing a feature mapping, distributed matching, dual-condition rule triggering, weighted fusion and confidence-driven update mechanism. The presentation layer 11 is used to provide users with an information interaction entry point, receive and display processing results from the business layer 12, and send task requests to the business layer 12. The presentation layer 11 includes a user management module, a human-computer interaction module, an information display module, a fault diagnosis module, and an inference rule module. Specifically, in the fault diagnosis module of the presentation layer in this application embodiment, fault diagnosis and risk prediction are performed using statistical characteristics based on rule base and fault information, and the final risk level is calculated by statistical values ​​and loss values. In the step of calculating the final risk level, the current date is obtained, and the difference between the current date and the last maintenance time for each type of equipment is calculated. The percentile of the query result in that type of equipment is used as the statistical value for that type of equipment. In one implementation, the loss value is the percentile of the difference between the current date and the equipment issuance date for all equipment.

[0020] Furthermore, in this embodiment of the application, the presentation layer 11 is a user interface. In this embodiment, the presentation layer 11 provides users with standardized interfaces for querying and displaying equipment quality data; fault diagnosis reasoning and display interfaces; information conversion interfaces; data import and export interfaces; reasoning rule interfaces; and user management interfaces, etc. In this embodiment, the presentation layer 11 provides users with an intuitive interface through a human-computer interaction module and an information display module, making it easy for users to quickly obtain and utilize device resource information.

[0021] Service layer 13 ensures fast data access and persistence through its data query and storage modules, significantly improving the efficiency of data sharing and reuse. The information display module of the presentation layer 13 in this embodiment of the application displays the information visually according to the query conditions input by the user, and supports querying, adding, updating and deleting information, thereby improving the user's interactive experience. Furthermore, in the business layer 12, the business layer 12 is used to receive task requests from the presentation layer, decompose the task requests into business requirements, and send the business requirements to the service layer, including a requirement decomposition module, a function combination module, an information reading module, and a case analysis module. The business layer 12 of this invention can decompose the business requirements proposed by the presentation layer and distribute them to the service layer to realize different combinations of software functions and meet different user operation needs, such as fault diagnosis, equipment management, and maintenance support information query; read relevant information such as equipment quality ontology from the database, and establish a fault diagnosis case analysis module according to the inference rules of the inference engine to realize the automatic fault diagnosis function; In the service layer 13, the service layer 13 is used to perform data query and storage based on business needs, and to complete data analysis, ontology construction, rule reasoning and data management, including a data query and storage module and an inference engine module; The inference engine module of service layer 13 of this invention performs semantic reasoning and analysis based on ontology models and rules, enabling users to provide more in-depth analysis of equipment resource information. Based on the construction of a domain knowledge ontology, and according to the provided standardized knowledge description, ontology reasoning rules are established to reason about knowledge semantics. The ontology reasoning rules are defined, and the inference engine function is implemented using software based on these defined rules. In the fault reasoning process, this invention applies a set of rules to the knowledge base, constructs a data model by analyzing the statistical characteristics of historical fault information, and simultaneously considers rule-based reasoning and case-based reasoning schemes to improve the accuracy of reasoning. In the data layer 14, the data layer 14 exchanges data with the service layer 13 to build a database, store, query and process standardized equipment quality data based on ontology, including a data management module; the data layer 14 mainly includes the storage of data files such as OWL and RDF, and provides a database access interface to realize the access and processing of data; The information display module is used to visualize information that matches the query conditions entered by the user during interaction, and to query, add, update and delete the matched information.

[0022] Furthermore, the user management module in the presentation layer of this invention includes management functions such as user registration, user deregistration, and user permission configuration. Account login is encrypted and has a login verification function, distinguishing between administrator and general user permissions. Administrator users can use all functions of the software platform, while general users cannot use inference rules and system management modules. For network security, this invention places the login verification implementation on the backend server. The frontend only needs to provide the username and password, which will be matched by the backend. The returned result only includes whether there is a match and the user's permissions. User registration, deregistration, permission configuration, and list display are implemented through CRUD operations in the SQL database.

[0023] Furthermore, in the information display module of this embodiment of the invention, the information displayed visually includes equipment information, test information, maintenance information, and fault information; wherein: The equipment information provided in this embodiment of the invention is used to describe the basic composition information of the equipment, including equipment model, equipment name, equipment number, equipment composition, assembly information, production information and equipment distribution date; The test information provided in this embodiment of the invention includes the test object, test items, test personnel, test results, test instruments, and test time; The maintenance information provided in this embodiment of the invention includes equipment maintenance records and maintenance manual information. The equipment maintenance records include the maintenance object, maintenance method, maintenance content, and maintenance date. The fault information provided in this embodiment of the invention is used to describe the historical fault data of the equipment, and is divided into two parts: fault category and maintenance category. The fault category includes fault type, fault phenomenon, fault location, fault cause, and fault time information; the maintenance category includes maintenance method, maintenance personnel, and maintenance time information.

[0024] Correspondingly, the information display module of this invention includes a device information module, a test information module, a maintenance information module, and a fault diagnosis module, wherein: The equipment information module visualizes equipment information data that meets certain query conditions. It can perform operations such as querying, adding, updating and deleting equipment information. At the same time, it can realize standardized description conversion and import of data from different source formats, which mainly include text documents, Access databases and Excel documents. In one implementation, the test information module visualizes the test information data that meets the requirements based on certain query conditions. It can perform operations such as querying, adding, updating and deleting test information, and at the same time realize the standardized description conversion and import of data from different source formats, which mainly include text documents, Access databases and Excel format documents. In one implementation, the maintenance information module includes two modules: maintenance records and maintenance manuals. It can visualize maintenance information data that meets the requirements based on certain query conditions, and can perform operations such as querying, adding, updating, and deleting maintenance information. At the same time, it can realize standardized description conversion and import of data from different source formats, which mainly include text documents, Access databases, and Excel format documents. In one implementation, the maintenance information module primarily performs operations on the SQL database, providing import, export, and clear functions. In the import function, the front-end sends a .sql file to the back-end in binary format. The back-end receives the file, connects to the existing SQL database, reads the binary file, splits it into multiple SQL statements, and executes them. In the export function, the back-end receives the request, writes the database table structure and data into SQL scripts to generate .sql files, and then sends them to the front-end. In the clear function, upon receiving data, it directly deletes each table in the connected database one by one. In one implementation, the fault diagnosis module uses certain observable fault information as diagnostic conditions and, based on the provided fault diagnosis method, realizes the prediction, diagnosis, repair, and recording of relevant information of the fault.

[0025] Furthermore, in the service layer, the present invention uses the Protege tool for ontology modeling, wherein a complete ontology includes classes, object properties, data properties, and individuals, and the exported ontology is saved in OWL format; Taking device information as an example, the present invention provides an ontology construction process that includes the following steps: Create classes for single or complex structures, where: In a single-structure scenario, create a root class "Equipment". Specifically, create subclasses under "Equipment": Equipment Model, Equipment Name, Equipment Number, Equipment Composition, Assembly Information, Production Information, and Equipment Issuance Date. Additionally, define corresponding data attributes for each class, such as: Equipment Model: string; Equipment Name: string; Equipment Number: string; Equipment Issuance Date: date. Create object attributes to represent relationships between classes, such as: hasComponent: from "Equipment" to "Equipment Composition"; hasAssemblyInfo: from "Equipment" to "Assembly Information"; hasManufacturerInfo: from "Equipment" to "Production Information". Additionally, in cases of complex structures, create classes for the complex structures, for example: Assembly Information: Manufacturer Name: string; Contact Information: string; Production Information: Manufacturer Name: string; Contact Information: string; Define data attributes for complex structures, for example: The "Manufacturer Name" and "Contact Information" attributes in the "Assembly Information" category; The "Manufacturer Name" and "Contact Information" attributes in the "Production Information" category; Create an instance for each class that is created, for example: Example of "device": Device A; Example of "equipment composition": Component 1, Component 2; Example of "Assembly Information": Assembly Information A; Example of “Production Information”: Production Information A.

[0026] Furthermore, in the service layer, rule-based reasoning performs knowledge reasoning for the device quality information resource ontology, including the following steps: Model the ontology of equipment quality information resources to achieve a standardized description of equipment quality information resources; export and save the completed ontology project as an OWL format file; The ontology reasoning process is implemented based on OWL files. During the reasoning process, the OWL ontology language operation interface and ontology reasoning machine tools provided by Jena reasoning machine are used to perform ontology knowledge reasoning, and to realize fault reasoning based on case base and rule base. Based on the reasoning results, the ontology of equipment quality information resources is updated.

[0027] Furthermore, this application embodiment introduces a mechanism for eliminating irrelevant features to reduce the noise of the ontology model constructed in this application embodiment, so as to improve the accuracy of the ontology model in subsequent equipment fault reasoning and diagnosis. The process of removing irrelevant features is represented as follows: ; ; Where n≤m, and B represents the original feature set. This represents the feature set after removing irrelevant features; m represents the number of features in the original feature set, and n represents the number of features in the feature set after removing irrelevant features.

[0028] In one embodiment, please refer to Figure 2 The embodiments of this application employ a feature filtering model to perform data feature filtering steps; Specifically, the following steps are included: S101. Extract equipment fault data containing features and fault tags from the equipment quality information resource ontology, and combine it with historical data from the fault diagnosis module of the presentation layer to organize the equipment fault data into transaction form to obtain an equipment fault transaction set. Each transaction contains equipment feature items, fault tags and auxiliary diagnostic parameters. Specifically, in step S101 of this embodiment, the device fault data is organized into a transaction form, and each transaction contains a set of device characteristics and corresponding fault labels; wherein, a transaction may represent multiple feature values ​​and their corresponding fault types recorded during a device detection. In one implementation, the data is cleaned and standardized to remove missing values ​​and noisy data, and the feature values ​​are transformed into a discrete data form suitable for the Apriori algorithm. Continuous numerical features can be binned and divided into several intervals to form discrete feature terms. The steps for data feature selection using a feature selection model also include: S102. Generate frequent itemsets layer by layer; specifically, determine a minimum support threshold to filter frequent itemsets, where support represents the frequency of a feature itemset in all transactions; starting with a single feature item, count the number of times the feature item appears in all transactions in a distributed parallel manner, calculate the feature item support, and retain feature items with support greater than or equal to the minimum support threshold as frequent itemsets; repeat this process until no new frequent itemsets can be generated; Preferably, in the step of generating frequent itemsets layer by layer: The device failure transaction set is horizontally divided into n data subsets {D1, D2, ..., D...} according to the service layer load balancing strategy. n Distribute n data subsets to m computing nodes, where m ≤ n, to build a distributed parallel mining environment; Each computing node has a local subset D assigned to it. i Execute the Map function; where D is scanned. i Extract feature terms and statistically analyze feature term t in Di Number of times in Output intermediate key-value pairs: Key=t, Value= Combined with local minimum support threshold Calculate local support: In the formula, Represents the local subset D i Total number of transactions; Each computing node executes the Combiner function to aggregate the occurrence counts of the same feature term t. , obtain feature t in D i Total number of occurrences Filter to meet The feature terms are used to generate a local 1-frequent itemset for node i. , ; Use the Reduce function to aggregate all nodes The number of times the statistical characteristic term t appears in the total number of transactions. ,in, In the formula, m is the total number of calculation nodes; Furthermore, the embodiments of this application incorporate a global minimum support threshold. Calculate global support ,in, Represents the total number of all transactions; filters that meet the criteria. Feature terms, generate global k-frequent itemsets , ; Global (k-1) frequent itemsets Distributed to each computing node, each node has a local subset D i Perform a self-join operation to generate k-itemset candidates. Combined with ontology association rule constraints, filtering Unreasonable itemsets are removed using Apriori properties. An itemset containing infrequent (k-1) subsets; traversing each node D i Statistical analysis of local support of the candidate set after pruning Global support is calculated through Reduce aggregation. ,reserve Itemsets, generate global k-frequent itemsets ; When the iteration meets the preset conditions, the iteration stops and the global frequent itemset set is output. .

[0029] In a preferred embodiment, the preset condition for stopping iteration is: it is impossible to pass. Generate satisfying Stop iteration when a new candidate set is found; or stop iteration when the inference module determines that "the current frequent itemset has covered all the feature associations required for social security fault diagnosis".

[0030] Furthermore, in one implementation of this application, the step of using a feature selection model for data feature selection also includes: S103, Association Rule Mining and Filtering; In step S103, a minimum confidence threshold is determined. The minimum confidence threshold is used to filter valid association rules. The confidence level represents the probability that a transaction containing feature set X also contains feature set Y. S104. For each frequent itemset, generate all possible association rules and calculate their confidence scores. Retain association rules with confidence scores greater than or equal to the minimum confidence threshold. Generate all possible association rules and filter them. S105. Based on the mined frequent itemsets and association rules, identify data features that have a significant impact on equipment fault reasoning and diagnosis.

[0031] For example, a minimum confidence threshold is determined. The minimum confidence threshold is used to filter valid association rules. The confidence level represents the probability that a transaction containing feature itemset X also contains feature itemset Y. If the minimum confidence threshold is set to 0.6, it means that in a transaction containing X, there is at least a 60% probability that Y is also contained. For each frequent itemset, generate all possible association rules and calculate their confidence scores. Retain association rules with confidence scores greater than or equal to the minimum confidence threshold. For example, for the frequent itemset {A,D,F}, the following association rules can be generated: Calculate the confidence level of the rule. If the confidence level is greater than or equal to the minimum confidence threshold, then retain the rule, indicating that there is a high probability that transactions containing A and D also contain F. Similarly, calculate the confidence level and perform filtering; And so on, generating all possible association rules and filtering them.

[0032] Finally, in this embodiment, based on the mined frequent itemsets and association rules, features that have a significant impact on equipment fault reasoning and diagnosis are identified. For example, if a feature frequently appears in a frequent itemset, or serves as an antecedent or consequent of an association rule and has a strong indicative effect on the fault, then that feature is considered important. The frequency of each feature's occurrence in the frequent itemsets and association rules can be counted, or the importance of the feature can be evaluated based on indicators such as the confidence and lift of the association rules. Furthermore, based on the importance of the features, the n most important features (n≤m, where m is the original number of features) are selected to form a new feature subset for subsequent equipment fault reasoning and diagnosis model construction. For example, if the original feature set has 10 features, the Apriori algorithm is used to select the 5 most important features as a new feature subset.

[0033] Preferably, in the association rule mining and filtering step, a lift is introduced to filter association rules, and the lift is expressed as: ;in, This represents the lifting degree of feature itemset X to feature itemset Y. It represents the confidence level of feature itemset X with respect to feature itemset Y, and is used to measure the probability that feature itemset Y will also appear when feature itemset X appears. Indicates the support of the feature set Y; when A value greater than 1 indicates that feature itemset X has a positive correlation with feature itemset Y, and it is preferred to retain it. Association rules with a value greater than 1.

[0034] This application utilizes the Apriori algorithm for feature selection, which extracts valuable features for fault reasoning and diagnosis from equipment fault data, eliminates irrelevant features, reduces noise in the ontology model, and improves the accuracy and efficiency of subsequent equipment fault reasoning and diagnosis.

[0035] Therefore, in this embodiment of the application, by setting a minimum support S and a minimum confidence C, if the frequency of feature set X in feature library D is not less than the support S, then feature set X is a frequent feature set; then, the feature library D is traversed to obtain all candidate feature sets and their corresponding support, and the infrequent feature sets with support less than the minimum support S are removed to obtain the feature set. Furthermore, in the process of modeling equipment quality information resources, firstly, a feature selection model is used to select features from the equipment quality information resources. Selecting the most diagnostically valuable information features can reduce the information dimensionality to a certain extent and is beneficial to the subsequent process of using inference rules to realize equipment fault diagnosis.

[0036] Furthermore, the steps for implementing the ontology reasoning process based on the OWL file specifically include: reading the OWL file of the created device quality information resource ontology through the Jena inference engine; registering and creating an inference engine according to the rules of the device quality information resource using the API provided by the Jena inference engine; binding the device quality information resource ontology and the inference engine together to generate a model object; and performing reasoning operations on the device quality information resource ontology through the model object by calling the Ontology API and Model API to obtain the reasoning results.

[0037] Therefore, after the equipment quality information data of this invention is standardized and described, the Jena ontology inference engine is used to infer fault cases; Jena mainly completes the import and persistence of the OWL model to the MySQL database; using Jena's inference function, based on user-defined inference rules, fault inference diagnosis based on fault cases is realized.

[0038] Specifically, the import and persistence of the model involves: creating an OWL ontology using Protege; when saving the OWL ontology, selecting RDF / XML as the LANGUAGE file and encoding it in UTF-8; then using Jena to read the model data from the OWL file and storing it in a MySQL database; after executing the program, the ontology is stored in the MySQL database. The database will generate the following tables, as shown in Table 1 below: Table 1. Ontology Data Table

[0039] The main data of the subject of this invention is stored in the following two tables: ① jena_g1t1_stmt, which stores the data information of the subject; ② jena_sys_stmt, which stores the metadata information of the subject.

[0040] Furthermore, this invention constructs inference rules. Specifically, based on the established equipment quality information resource ontology, inference rules are constructed. The construction of inference rules follows these steps: ① The user fills in and uploads the rule header, rule body, and rule footer respectively; ② The predicate form is constructed according to a unified structure; ③ Inference rules are designed based on the constructed predicate form.

[0041] Furthermore, the step of updating the equipment quality information resource ontology based on the reasoning results includes: Obtain the inference results using the API provided by Jena; Based on the inference results, the device quality information resource ontology is updated using Jena's API. The update operations include: adding new information: adding new information from the inference results to the ontology; modifying existing information: correcting inaccurate information in the ontology based on the inference results; deleting redundant information: removing information that is no longer applicable or redundant from the ontology; and saving the updated ontology: re-exporting the updated device quality information resource ontology as an OWL format file.

[0042] Furthermore, based on the construction and updating of the basic ontology, this application introduces an update mechanism that uses feature mapping, distributed matching, dual-condition rule triggering, weighted fusion, and confidence-driven updates to dynamically reason and adaptively update the ontology model knowledge associated with device quality features. Specifically, this step employs a three-level progressive reasoning architecture to achieve ontology knowledge reasoning, wherein: The first level is distributed feature matching inference; specifically, based on the equipment quality features selected by the improved Apriori algorithm, the mapping relationship between feature vectors and ontology concepts is constructed through the OWL interface of the Jena inference engine; the distributed computing framework is used to match the real-time operation data of the equipment with the preset fault feature templates in the ontology in parallel to generate preliminary matching results. The second level is hybrid rule reasoning; specifically, SWRL rules are combined with statistical characteristic indicators to construct a dual-condition mechanism (feature threshold + rule trigger). When the percentile of the statistical value of a certain feature in the distributed matching result exceeds the preset threshold, the Drools inference engine is activated to call the corresponding SWRL rule set and output the intermediate inference conclusion. The third level is dynamic weighted fusion reasoning. Specifically, the random forest algorithm is introduced to calculate the confidence weight of different reasoning paths. The distributed matching results and rule-based reasoning conclusions are weighted and fused. If the overall confidence after fusion is ≥0.85, the ontology knowledge is directly updated. If the confidence is between 0.7 and 0.85, the human-computer interaction module is triggered for manual confirmation before updating. If the confidence is <0.7, the initial matching result is generated again, that is, the process returns to the first level for rematching.

[0043] In the implementation steps of combining SWRL rules with statistical characteristic indicators, based on statistical values ​​(percentile of the difference between the current date and the last maintenance time) and loss values ​​(percentile of the difference between the current date and the equipment issuance date), they are transformed into quantitative condition parameters in SWRL rules, specifically divided into three levels of thresholds: statistical value thresholds: low (0%–30%), medium (31%–70%), and high (71%–100%); loss value thresholds: low (0%–40%), medium (41%–60%), and high (61%–100%). These thresholds are determined through percentile distribution analysis of equipment historical maintenance records and fault data, serving as the quantitative basis for rule triggering. Furthermore, a dual-condition structure of feature attributes and statistical indicators is used to construct SWRL rules. When real-time operating data of the equipment is input, the percentiles of the current statistical value and loss value are first calculated, and then compared with the preset thresholds through the SWRL rules: if a single indicator exceeds the threshold (e.g., the statistical value is high but the loss value is low), the basic handling rule is triggered (e.g., "prioritize maintenance plan"); if both indicators reach the high threshold, the upgraded handling measures in the composite rule are activated, and the "equipment risk level" instance in the ontology is updated synchronously (e.g., "medium risk" is updated to "high risk"); after the rule is executed, the frequency of association between statistical indicators and fault types during the inference process is automatically recorded for subsequent dynamic calibration of the threshold (e.g., when the trigger frequency of a certain type of fault in the statistical value range exceeds 60%, the threshold of that range is adjusted to 31%~75%). The results of rule reasoning are verified by combining case similarity calculation methods: when the similarity between the reasoned emergency measures and the handling solutions under the same combination of statistical indicators in historical cases is ≥0.85, they are directly adopted; when the similarity is between 0.7 and 0.85, the rule base is updated after manual correction; when the similarity is <0.7, they are marked as candidate cases for new rules and added to the rule set through association analysis.

[0044] Furthermore, in the step of combining SWRL rules with statistical characteristic indicators to construct a two-condition mechanism in this embodiment of the application: The dual-condition mechanism includes the coordinated triggering of rule conditions and statistical threshold conditions. The rule conditions are constructed based on the object attributes and data attributes in the equipment quality information resource ontology, and adopt the form of SWRL rules. The antecedent includes ontology concept association, and the consequent corresponds to the preset diagnostic conclusion. The statistical threshold conditions are a combination threshold of statistical value percentile and loss value percentile. Both statistical value percentile and loss value percentile are divided into three levels: high, medium and low. The two threshold levels are determined by the distribution characteristics of historical fault data. When the ontology concepts in the rule conditions are completely matched and at least one indicator in the statistical threshold conditions reaches a high level, the dual-condition mechanism is triggered, activating the corresponding SWRL rule set. If the rule conditions are matched but the statistical threshold conditions only meet the medium level, the simplified rule set is triggered; if the statistical threshold conditions are lower than the medium level, the rule is not triggered for the time being, and the matching result is only recorded for subsequent threshold calibration. After the rule set is triggered, the specific values ​​of the statistical threshold conditions are embedded into the reasoning process through the ontology inference engine tool to generate a structured conclusion containing a triple of fault characteristics, statistical indicators, and handling measures. The statistical indicator values ​​are synchronously updated to the corresponding data attribute fields of the equipment quality information resource ontology, realizing the dynamic association between rule reasoning and ontology instances.

[0045] In one implementation of this application, a training dataset for general electrical equipment is constructed, such as by collecting historical fault data of electrical equipment from multiple industries. Each case is decomposed into input features and labels. The input features include inference path conclusions and equipment and environment features. The inference path conclusions include distributed matching conclusions (e.g., "contact oxidation") and rule-based inference conclusions (e.g., "preventive maintenance"). The equipment and environment features include equipment type (switch cabinet / motor), operating environment (temperature / humidity), and statistical values ​​(percentile of fault occurrences, percentile of maintenance cycle). The labels are used to indicate whether the inference path conclusion is correct (1 = correct, 0 = incorrect). The random forest model is trained by using "device type, operating environment, and statistical values" as auxiliary features and "distributed matching conclusions and rule-based reasoning conclusions" as core features to construct a general feature vector. The random forest (with 80 decision trees) is trained using cross-scenario samples, and the model can output dynamic path weights. Taking switchgear temperature anomaly as an example, real-time inference and weighted fusion are performed. In distributed matching, the input is real-time switchgear data (A-phase temperature = 88℃, number of temperature exceedances in the past week = 3 times); the matching involves calling the improved Apriori mining of frequent itemsets. Based on the improved Apriori algorithm, frequent itemsets are mined, and feature matching is performed on the real-time data through the distributed matching framework. If the temperature in the real-time data is >85℃, the matching association knowledge inference deduces the fault as contact oxidation; the output conclusion A = "contact oxidation fault, confidence level 0.7 (historical matching accuracy)". The input for rule inference is the statistical attributes of the equipment instances in the ontology (percentile of fault occurrences in the past 3 months = 75%); the trigger for rule inference is the SWRL rule. According to the defined SWRL rule logic, when the equipment fault type is identified as temperature anomaly, and its percentile of fault occurrences in the past 3 months is >70%, preventive maintenance operations will be triggered according to the inference logic set by the rule, that is, simultaneously satisfying "fault type is temperature anomaly, and fault occurrence percentile >70%". Under both conditions, perform preventative maintenance. Specifically, output conclusion B = "Preventative maintenance required, confidence level 0.8 (accuracy of rule coverage cases)". In the weighted random forest algorithm, the input is a feature vector {equipment type = switch cabinet, fault percentile = 75%, conclusion A = contact oxidation, conclusion B = preventive maintenance}; the output is the path weight W. A =0.5, W B =0.5 (The model learned that "in scenarios with a high number of failures, the confidence levels of the two paths are comparable"); Fusion conclusion: Overall confidence level = 0.7×0.5+0.8×0.5=0.75, output "Preventive maintenance is recommended (contact oxidation or cooling system issues need to be investigated, and further confirmation should be made in conjunction with the ontology update)"; Furthermore, the ontology is updated and the model is iterated. If the actual fault is confirmed by manual inspection as "cooling system blockage", the training dataset is updated: new samples are added {Conclusion A = contact oxidation, Conclusion B = preventive maintenance, actual fault = cooling system blockage}; the model is retrained, and the path weight rules under the "switch cabinet + high fault frequency" scenario are updated (such as adjusting the rule inference weight to 0.6); a "cooling system blockage" association rule is added under the "temperature abnormal fault" category, and the constraint condition "summer high temperature scenario needs to be judged in combination with humidity characteristics" is added.

[0046] In the fault diagnosis module of this embodiment, the present application uses rule-based reasoning and case-based reasoning methods for fault diagnosis and prediction. Reasoning rules are used to achieve fault diagnosis and risk prediction based on the rule base and the statistical characteristics of fault information. Based on a standardized fault data structure, and relying on key search conditions (equipment model, equipment name, fault location, fault phenomenon), fault cases are returned in descending order of relevance. Based on the equipment model, name, and maintenance date, the risk rate of the equipment is predicted from a statistical perspective, enabling early warning and timely maintenance of high-risk equipment. In the rule base mode provided by this invention, it can also be a user-defined rule base, including rule header, rule body, and rule footer, which is used for semantic analysis in Jena. In the fault diagnosis process, the user inputs relevant query fields, analyzes relevant rules and retrieves fault information, filters out query results that meet the conditions, and completes fault diagnosis. In fault risk prediction, this invention is based on the statistical characteristics of historical fault information.

[0047] For details, please refer to Figure 3 The fault diagnosis method of the present invention includes the following steps: S201. Establish a risk database, which includes data types, equipment models, equipment names, maintenance times, and failure times. Specifically, in the data types section of the risk database, each import will automatically fill in the information based on whether the data originates from maintenance data or fault data. If the data source is maintenance data, "maintenance" will be entered; if the data source is fault data, "fault" will be entered. In the equipment model section of the risk database, the corresponding equipment model will be updated synchronously during data import. In the equipment name section of the risk database, the corresponding equipment name will be updated synchronously during data import. In the maintenance time section of the risk database, the corresponding maintenance time will be updated synchronously during data import, and it will be left blank if it is fault information. In the fault time section of the risk database, the corresponding fault time will be updated synchronously during data import, and it will be left blank if it is maintenance information.

[0048] S202. Classify the equipment according to its model and name, and statistically analyze the historical fault information of the equipment based on the classification results. The statistical information includes the equipment's fault frequency, maintenance frequency, and maintenance-failure time difference. In this invention, equipment is distinguished by both equipment model and equipment name, and equipment is considered to be of the same type only when both equipment model and equipment name are the same. S203. For each type of equipment, query the time of the last maintenance before the failure according to the failure time in each failure data. If there is no maintenance data before the failure, the maintenance time is equal to the failure time. Calculate the difference between the two and sort the differences in ascending order. The sorted results are stored in the database for later use. S204. In response to the user's click operation that triggers risk prediction behavior in the interactive interface, obtain the current date, calculate the difference between the current date and the last maintenance time of each type of equipment, and use the percentile of the query result in that type of equipment as the statistical value t for that type of equipment. Calculate the loss value s, which is the percentile of the difference between the current date and the equipment allocation date among all equipment for that difference. Calculate the final risk level using the statistical value t and the loss value s. Among them, the coefficients of the t-value and s-value obtained by fitting historical data are 0.8 and 0.2 respectively, so the risk level is 0.8*t+0.2*s.

[0049] Furthermore, in the fault diagnosis module of the present invention, the fault diagnosis module has an import submodule, a query submodule, a diagnosis submodule, and a prediction submodule; wherein: In the import submodule provided in this embodiment, the front end filters the source file format and uploads the file. The file is also transmitted in binary format. Different upload links are distinguished according to the differences in the source file format. The back end receives the file content and returns the reading result to the front end for preview. After checking the data, the file content can be uploaded to the database through the import command. The back end server has a fault information table of different source formats. When it is necessary to export the template, the existing template can be downloaded to the local machine by sending a request to the server. In the query submodule provided in this embodiment, the query submodule is used to display all fault information in the database and perform pagination processing. It can respond to user commands to modify and delete individual information entries. The present invention also provides a data export function in the query submodule, which can export the fault information table in the database as an Excel spreadsheet and download it to the local machine. In the diagnostic submodule provided in this embodiment, the equipment model, equipment name, fault phenomenon, and fault location can be uploaded according to user needs. The backend uses a rule-based fault reasoning method to perform fault reasoning and returns the obtained results to the frontend in order of probability from high to low. At the same time, the fault type can be filtered on the frontend for further accurate diagnosis. In the prediction submodule provided in this embodiment, risk prediction can be performed on equipment already in the database. The web page can personalize the risk level of equipment by adjusting the risk threshold, and different risk levels can be filtered on the front end to make the display results more intuitive.

[0050] In summary, the information sharing platform of this application utilizes ontology technology to standardize data resources during equipment use and maintenance, solves the problem of inconsistent data resource structure and semantic description in the field of quality monitoring, enables the querying and utilization of the data resources, and effectively improves the sharing and reuse efficiency of equipment use quality data resources.

[0051] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0052] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. It can be applied to various fields suitable for the present invention. Other modifications can be readily made by those skilled in the art. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and examples shown and described herein.

Claims

1. A platform for sharing equipment resource information for intelligent diagnostics, characterized in that: include: The business layer receives task requests from the presentation layer, decomposes these requests into business requirements, and sends these requirements to the service layer. The service layer then performs data querying and storage based on these requirements, completing data analysis, ontology construction, rule reasoning, and data management. Rule reasoning involves knowledge reasoning for the equipment quality information resource ontology, including: using a feature filtering model to filter data features (the Apriori model based on distributed frequent itemset mining and association rule lifting); modeling the equipment quality information resource ontology based on the filtered data features; and dynamically reasoning and adaptively updating the ontology model knowledge associated with the equipment quality features through feature mapping, distributed matching, dual-condition rule triggering, weighted fusion, and confidence-driven update mechanisms. The presentation layer is used to provide users with an information interaction entry point, receive and display processing results from the business layer, and send task requests to the business layer. The presentation layer includes a user management module, a human-computer interaction module, an information display module, a fault diagnosis module, and an inference rule module. In the fault diagnosis module of the presentation layer, fault diagnosis and risk prediction are performed based on the statistical characteristics of the rule base and fault information. The final risk level is calculated by using statistical values ​​and loss values. Specifically, the current date is obtained, and the difference between the current date and the last maintenance time for each type of equipment is calculated. The percentile of the query result for that type of equipment is used as the statistical value for that type of equipment. The loss value is the percentile of the difference between the current date and the equipment issuance date for all equipment.

2. The equipment resource information ontology sharing platform for intelligent diagnostics according to claim 1, characterized in that, The steps for data feature selection using a feature selection model include: From the equipment quality information resource ontology, equipment fault data containing features and fault tags are extracted, and combined with historical data from the fault diagnosis module of the presentation layer, the equipment fault data is organized into transaction form to obtain equipment fault transaction set. Each transaction contains equipment feature items, fault tags and auxiliary diagnostic parameters. Generating frequent itemsets layer by layer: Determine a minimum support threshold to filter frequent itemsets, where support represents the frequency of a feature itemset in all transactions; starting with a single feature item, count the number of times the feature item appears in all transactions in a distributed parallel manner, calculate the feature item support, and retain feature items with support greater than or equal to the minimum support threshold as frequent itemsets; repeat this process until no new frequent itemsets can be generated. Association rule mining and filtering: Determine a minimum confidence threshold, which is used to filter valid association rules. The confidence level represents the probability that a transaction containing feature itemset X also contains feature itemset Y. For each frequent itemset, generate all possible association rules and calculate the confidence level of the frequent itemset. Retain association rules with a confidence level greater than or equal to the minimum confidence threshold. Generate all possible association rules and filter them. Based on the frequently mined itemsets and association rules, data features that have a significant impact on equipment fault reasoning and diagnosis are identified.

3. The equipment resource information ontology sharing platform for intelligent diagnostics according to claim 2, characterized in that, In the step of generating frequent itemsets level by level: The device failure transaction set is horizontally divided into n data subsets {D1, D2, ..., D...} according to the service layer load balancing strategy. n Distribute n data subsets to m computing nodes, where m ≤ n, to build a distributed parallel mining environment; Each computing node has a local subset D assigned to it. i Execute the Map function; where D is scanned. i Extract feature terms and statistically analyze feature term t in D i Number of times in Output intermediate key-value pairs: Key=t, Value= Combined with local minimum support threshold Calculate local support: In the formula, Represents the local subset D i Total number of transactions; Each computing node executes the Combiner function to aggregate the occurrence counts of the same feature term t. , obtain feature t in D i Total number of occurrences Filter to meet The feature terms are used to generate a local 1-frequent itemset for node i. , ; Use the Reduce function to aggregate all nodes The number of times the statistical characteristic term t appears in the total number of transactions. ,in, In the formula, m is the total number of computation nodes; combined with the global minimum support threshold. Calculate global support ,in, Represents the total number of all transactions; filters that meet the criteria. Feature terms, generate global k-frequent itemsets , ; Global (k-1) frequent itemsets Distributed to each computing node, each node has a local subset D i Perform a self-join operation to generate k-itemset candidates. Combined with ontology association rule constraints, filtering Unreasonable itemsets are removed using Apriori properties. An itemset containing infrequent (k-1) subsets; traversing each node D i Statistical analysis of local support of the candidate set after pruning Global support is calculated through Reduce aggregation. ,reserve Itemsets, generate global k-frequent itemsets ; When the iteration meets the preset conditions, the iteration stops and the global frequent itemset set is output. .

4. The equipment resource information ontology sharing platform for intelligent diagnostics according to claim 3, characterized in that, In the steps of association rule mining and filtering, a lift is introduced to filter association rules. The lift is expressed as: ;in, This represents the lifting degree of feature itemset X to feature itemset Y. It represents the confidence level of feature itemset X with respect to feature itemset Y, and is used to measure the probability that feature itemset Y will also appear when feature itemset X appears. Indicates the support of the feature set Y; when A value greater than 1 indicates that feature itemset X has a positive correlation with feature itemset Y, and it is preferred to retain it. Association rules with a value greater than 1.

5. The equipment resource information ontology sharing platform for intelligent diagnostics according to claim 4, characterized in that, In the steps of dynamically reasoning and adaptively updating the ontology model knowledge associated with device quality features through the construction of feature mapping, distributed matching, bi-condition rule triggering, weighted fusion, and confidence-driven update mechanisms: Based on the equipment quality features selected by the improved Apriori algorithm, the mapping relationship between feature vectors and ontology concepts is constructed through the OWL interface of the Jena inference engine; the real-time operation data of the equipment is matched in parallel with the preset fault feature templates in the ontology using a distributed computing framework to generate preliminary matching results. By combining SWRL rules with statistical characteristic indicators, a two-condition mechanism is constructed. When the percentile of the statistical value of a certain feature in the distributed matching result exceeds the preset threshold, the Drools inference engine is activated to call the corresponding SWRL rule set and output the intermediate inference conclusion. The random forest algorithm is introduced to calculate the confidence weight of different inference paths. The distributed matching results and rule-based inference conclusions are weighted and fused. If the overall confidence of the fused results is ≥0.85, the ontology knowledge is updated directly. If the confidence is between 0.7 and 0.85, the human-computer interaction module is triggered for manual confirmation before updating. If the confidence is <0.7, the initial matching results are generated again.

6. The equipment resource information ontology sharing platform for intelligent diagnostics according to claim 5, characterized in that, In the steps of constructing a two-condition mechanism by combining SWRL rules with statistical property indicators: The dual-condition mechanism includes the coordinated triggering of rule conditions and statistical threshold conditions; The rule conditions are constructed based on the object attributes and data attributes in the equipment quality information resource ontology, and adopt the SWRL rule form. The antecedent includes ontology concept association, and the consequent corresponds to the preset diagnostic conclusion. The statistical threshold condition is a combination threshold of the percentile of the statistical value and the percentile of the loss value. Both the percentile of the statistical value and the percentile of the loss value are divided into three levels: high, medium and low. The two threshold levels are determined by the distribution characteristics of historical fault data. When the ontology concepts in the rule conditions are fully matched and at least one indicator in the statistical threshold conditions reaches a high level, the dual-condition mechanism is triggered, activating the corresponding SWRL rule set; if the rule conditions are matched but the statistical threshold conditions only meet the medium level, the simplified rule set is triggered; if the statistical threshold conditions are lower than the medium level, the rule is not triggered temporarily, and only the matching result is recorded for subsequent threshold calibration. After the rule set is triggered, the specific values ​​of the statistical threshold conditions are embedded into the reasoning process through the ontology inference engine tool to generate a structured conclusion containing a triple of fault characteristics, statistical indicators, and handling measures. The statistical indicator values ​​are synchronously updated to the corresponding data attribute fields of the equipment quality information resource ontology, realizing the dynamic association between rule reasoning and ontology instances.

7. The equipment resource information ontology sharing platform for intelligent diagnostics according to claim 6, characterized in that, The information display module is used to visualize information that matches the query conditions entered by the user in the interaction, and to query, add, update and delete the matched information; In the information display module, the information displayed visually includes equipment information, test information, maintenance information, and fault information; Equipment information describes the composition of the equipment, including equipment model, equipment name, equipment number, equipment composition, assembly information, production information, and equipment distribution date. Test information includes the object under test, test items, testers, test results, test instruments, and test time; Maintenance information includes equipment maintenance records and maintenance manual information. Equipment maintenance records include maintenance objects, maintenance methods, maintenance content, and maintenance dates. Fault information is used to describe the equipment's historical fault data, and is divided into two parts: fault category and maintenance category. The fault category includes fault type, fault phenomenon, fault location, fault cause and fault time information; the maintenance category includes maintenance method, maintenance personnel and maintenance time information.

8. The equipment resource information ontology sharing platform for intelligent diagnostics according to claim 7, characterized in that, In the service layer, a complete ontology includes classes, object relationships, data relationships, and instances; the ontology construction process includes the following steps: In a single-structure scenario, a root class is created, corresponding data attributes are defined for each class, and object attributes are created to represent the relationships between classes. In the case of complex structures, create classes for the complex structures and define data attributes for the complex structures; Create an instance for each class that is created.

9. The equipment resource information ontology sharing platform for intelligent diagnostics according to claim 8, characterized in that, In the step of performing knowledge reasoning for the equipment quality information resource ontology based on rule-based reasoning, the completed ontology project is exported as an OWL format file. The ontology reasoning process is then implemented based on the OWL format file, including: The OWL file of the created device quality information resource ontology is read using the Jena inference engine; the inference engine is registered and created according to the rules of the device quality information resource using the API provided by the Jena inference engine. The equipment quality information resource ontology and the inference engine are bound together to generate a model object; By calling the OntologyAPI and ModelAPI through the model object, inference operations are performed on the device quality information resource ontology to obtain the inference results. Obtain the inference results through the API provided by Jena; based on the inference results, update the device quality information resource ontology using the Jena API, where the update operations include adding new information, modifying existing information, deleting redundant information, and saving the updated ontology.

10. The equipment resource information ontology sharing platform for intelligent diagnostics according to claim 9, characterized in that, The fault diagnosis module includes the following steps for fault diagnosis: Establish a risk database, which includes data types, equipment models, equipment names, maintenance times, and failure times. The equipment is classified according to its model and name. Based on the classification results, the historical failure information of the equipment is statistically analyzed. The statistical information includes the failure frequency, maintenance frequency, and maintenance-failure time difference of the equipment. For each type of equipment, query the time of the last maintenance before the failure based on the failure time in each failure data. If there is no maintenance data before the failure, the maintenance time is equal to the failure time, the difference between the two is calculated, and the differences are sorted in ascending order. The sorted results are stored in the database for later use. Calculate the final risk level in response to a user's click that triggers a risk prediction action in the interactive interface.

Images