An industrial knowledge mining modeling method and system based on a knowledge graph
By constructing an industrial knowledge mining model based on knowledge graphs, and combining equipment manufacturing, control, and environmental information, similar feature comparison and twin model simulation are performed, the problems of accuracy and timeliness of industrial equipment fault early warning are solved, and efficient fault early warning and monitoring are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GONGYEYUN MFG (SICHUAN) INNOVATION CENT CO LTD
- Filing Date
- 2023-08-30
- Publication Date
- 2026-05-22
AI Technical Summary
Existing methods for early warning of industrial equipment failures have low accuracy and timeliness, and fail to fully consider factors such as equipment users and the environment, resulting in insufficient accuracy and timeliness of early warnings.
By constructing an industrial knowledge mining model based on knowledge graphs, information on equipment manufacturing, control, environment, and personnel is obtained. A target knowledge graph set is constructed, and similar features are compared. The target equipment twin model is then embedded to provide fault early warning.
It improves the accuracy and efficiency of industrial equipment fault early warning, timely detection of potential fault hazards, and avoids serious equipment failures from affecting production.
Smart Images
Figure CN117056533B_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to data mining techniques, and more specifically, to a knowledge graph-based industrial knowledge mining modeling method and system. Background Technology
[0002] Existing methods for early warning of industrial equipment failures typically involve judging the equipment's operating status and issuing a failure warning based on the judgment result. However, this method does not take into account other relevant factors, such as the equipment users and maintenance personnel, resulting in a low accuracy rate for equipment failure warnings. Furthermore, it is difficult to accurately predict equipment failures in the future based on the current operating status of the equipment, leading to a low timeliness of equipment failure warnings.
[0003] The shortcomings of existing industrial equipment fault early warning methods are that the accuracy and timeliness of fault early warning are relatively low. Summary of the Invention
[0004] Therefore, in order to solve the above-mentioned technical problems, the technical solutions adopted in the embodiments of this disclosure are as follows:
[0005] A knowledge graph-based industrial knowledge mining and modeling method includes the following steps: acquiring basic information of a pre-constructed knowledge graph, wherein the basic information includes the graph application scenario, graph application target, and graph construction width; determining the data acquisition type and data acquisition volume based on the basic information, wherein the data acquisition type includes equipment manufacturing parameters, equipment control parameters, equipment usage environment parameters, and participant information; collecting relevant data from multiple similar equipment products of the target equipment based on the data acquisition type and the data acquisition volume, and constructing multiple target knowledge graph sets based on the data acquisition results, wherein the target knowledge graphs correspond to equipment usage time nodes; traversing and comparing the multiple target knowledge graph sets based on the equipment usage time nodes to obtain a graph similarity feature set, wherein the graph similarity features correspond to equipment usage time nodes; constructing a target equipment twin model, and embedding the graph similarity features in the graph similarity feature set into the target equipment twin model according to the equipment usage time nodes; and providing fault warnings for the target equipment usage process based on the target equipment twin model.
[0006] A knowledge graph-based industrial knowledge mining and modeling system includes: a pre-built knowledge graph information acquisition module, which acquires basic information of the pre-built knowledge graph, including graph application scenarios, graph application goals, and graph construction width; a data acquisition information determination module, which determines the data acquisition type and data acquisition volume based on the basic information, wherein the data acquisition type includes equipment manufacturing parameters, equipment control parameters, equipment usage environment parameters, and participant information; and a target knowledge graph set construction module, which performs relevant data acquisition on multiple similar equipment products of the target equipment based on the data acquisition type and the data acquisition volume, and constructs a target knowledge graph set based on the data acquisition results. The system comprises: a knowledge graph set construction module, which constructs multiple target knowledge graph sets, each corresponding to a device usage time node; a graph similarity feature set acquisition module, which traverses and compares the multiple target knowledge graph sets based on the device usage time nodes to acquire graph similarity feature sets, wherein the graph similarity features correspond to the device usage time nodes; a target device twin model construction module, which constructs a target device twin model and embeds the graph similarity features from the graph similarity feature set into the target device twin model according to the device usage time nodes; and a target device fault early warning module, which provides fault early warning for the target device during its usage process based on the target device twin model.
[0007] Due to the adoption of the above-mentioned technical methods, the technical advancements achieved by this disclosure compared to the prior art are as follows:
[0008] (1) It solves the technical problem of low accuracy and timeliness of existing industrial equipment fault early warning methods. By constructing an industrial equipment twin model based on knowledge mining, and using the industrial equipment twin model to simulate and predict equipment faults based on the current equipment operating status, the accuracy and efficiency of industrial equipment fault early warning can be improved, thereby timely detection and handling of potential faults and avoiding serious equipment failures that could affect production.
[0009] (2) By constructing a knowledge graph of the target device, more potential content and relationships related to the target device can be found, thereby enabling accurate analysis and prediction of the target device's operational failures and improving the accuracy of device failure cause detection. Attached Figure Description
[0010] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments will be briefly introduced below.
[0011] Figure 1This application provides a flowchart illustrating the process of industrial knowledge mining and modeling based on knowledge graphs.
[0012] Figure 2 This application provides a flowchart illustrating the process of constructing multiple target knowledge graph sets in an industrial knowledge mining and modeling method based on knowledge graphs.
[0013] Figure 3 This application provides a schematic diagram of the structure of an industrial knowledge mining and modeling system based on knowledge graphs. Detailed Implementation
[0014] The technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort are within the scope of protection of this disclosure.
[0015] Based on the above description, please refer to Figure 1 ,like Figure 1 As shown, this disclosure provides a knowledge graph-based industrial knowledge mining and modeling method, including:
[0016] Obtain basic information about the pre-built knowledge graph, including graph application scenarios, graph application goals, and graph construction width;
[0017] Knowledge graphs are a series of various graphs that display the development process and structural relationships of knowledge. They describe knowledge resources and their carriers through data visualization technology, and mine, analyze, construct, draw, and display knowledge and their interrelationships. The method provided in this application is used to mine industrial knowledge by constructing an industrial equipment knowledge graph, build an industrial equipment twin model based on the industrial knowledge mining results, and finally monitor the operating status and provide fault warnings for industrial equipment based on the industrial equipment twin model, thereby improving the accuracy and timeliness of industrial equipment fault warnings. The method provided in this application is specifically implemented in an industrial knowledge mining and modeling system based on knowledge graphs.
[0018] First, the basic information of the pre-built knowledge graph is obtained. This basic information includes the graph's application scenario, application goal, and construction width. The graph's application scenario refers to the specific context in which the knowledge graph is built, such as a factory manufacturing workshop or assembly workshop. The graph's application goal refers to the problem that the knowledge graph aims to solve, such as industrial equipment failure or low product qualification rate. The construction width includes the scale and depth of the knowledge graph. The scale of the knowledge graph refers to the number of entities and relationships it contains; a larger scale results in greater coverage and richer knowledge. The depth of the knowledge graph refers to the hierarchical depth of the knowledge it contains. Some knowledge graphs only contain surface-level relationships, while others may delve into more detailed levels. Knowledge graphs with higher depth can provide more specific and accurate information.
[0019] By obtaining basic information from the pre-built knowledge graph, support was provided for determining the next step of data collection.
[0020] The data acquisition type and data acquisition volume are determined based on the basic information. The data acquisition type includes equipment manufacturing parameters, equipment control parameters, equipment operating environment parameters, and personnel information.
[0021] The data acquisition type is determined based on the map application scenario and map application objectives in the basic information. The data acquisition type includes equipment manufacturing parameters, equipment control parameters, equipment usage environment parameters, and personnel information. The equipment manufacturing parameters refer to the basic information of industrial equipment manufacturing, including industrial equipment type, brand, specifications, and manufacturing batch. The equipment control parameters refer to the control parameters during equipment operation, including operating speed, operating voltage, and operating current. The equipment usage environment parameters refer to the environment in which the equipment operates, including temperature and humidity. The personnel information refers to all personnel involved in the equipment use process, including equipment users, daily maintenance personnel, and cleaning personnel.
[0022] The data collection volume is determined based on the graph construction width in the basic information. The larger the scale and depth of the knowledge graph constructed within the graph construction width, the greater the data collection volume. Determining the data collection type and volume provides a basis for the next step of data collection.
[0023] Based on the data collection type and the data collection volume, relevant data are collected from multiple similar equipment products of the target device, and multiple target knowledge graph sets are constructed based on the data collection results, wherein the target knowledge graphs correspond to the device usage time nodes;
[0024] Based on the target device, extract historical operation log information of multiple similar equipment products when the device malfunctions. The similar equipment products refer to equipment that is exactly the same as the target device in terms of brand, type, and specifications and is produced in the same factory. Then, according to the data collection type and the data collection volume, extract relevant data from the historical operation log information when the device malfunctions, that is, extract the equipment manufacturing parameters, historical equipment control parameters, historical equipment usage environment parameters, and historical personnel information from the historical operation log information to obtain the data collection results.
[0025] like Figure 2 As shown, in one embodiment, it further includes:
[0026] Randomly select one device product from multiple similar device products as the first device product, and obtain the first data acquisition result of the first device product;
[0027] The first data acquisition result is subjected to data cleaning processing to obtain the first data cleaning processing result, wherein the data cleaning processing steps include missing value processing, outlier processing, data deduplication, and data format conversion;
[0028] Further, the first data cleaning and processing results are subjected to data standardization processing, and the standard data processing results are divided according to the equipment usage time nodes to obtain the first standard data set, wherein the first standard data and the equipment usage time nodes have a corresponding relationship;
[0029] First, randomly select one equipment product from among the multiple similar equipment products as the first equipment product. The first equipment product is any one of the multiple similar equipment products. Obtain the first data acquisition result of the first equipment product. The first data acquisition result includes the first equipment manufacturing parameters, the first historical equipment control parameters, the first historical equipment usage environment parameters, and the first historical participant information.
[0030] Then, the first data acquisition results are subjected to data cleaning processing. First, the first data acquisition results are deduplicated. Data deduplication aims to reduce duplicate and redundant data in the first data acquisition results, maintain data uniqueness, and improve data processing efficiency. Specific methods for data deduplication can be implemented using existing technologies and will not be elaborated here. Then, the first data acquisition results after data deduplication are processed for missing values and outliers. Missing value processing methods mainly include deletion, imputation, and interpolation; outlier processing methods include deleting outliers and replacing outliers. Those skilled in the art can select appropriate missing value processing methods and outlier processing methods based on the actual data processing volume.
[0031] Finally, the first data acquisition results, after missing value and outlier handling, undergo format conversion. This involves converting unstructured and semi-structured data in the first data acquisition results into structured data. The format conversion method can be implemented using existing technologies, resulting in a first data cleaning process that contains only structured data. By cleaning the first data acquisition results, data processing time can be reduced, and the completeness, uniqueness, and accuracy of the data acquisition results can be improved. Simultaneously, the accuracy of the target knowledge graph construction can be enhanced.
[0032] Further, the first data cleaning and processing results are subjected to data standardization. Data standardization refers to processing the data in the first data cleaning and processing results into a standard value range, that is, transforming the data in the first data cleaning and processing results into a specific range for easier comparison and calculation. Commonly used data standardization methods include data normalization and data mean standardization. An appropriate data standardization method can be selected according to the actual situation to obtain standard data processing results. Then, the standardized data processing results are divided according to the equipment usage time nodes. The equipment usage time nodes can be set according to the actual situation of equipment usage. Generally, equipment in a factory runs continuously around the clock; for example, the equipment usage time node can be set to one day. The standardized data of each day during equipment operation is obtained as the first standard data, constructing a first standard data set, where the first standard data and the equipment usage time nodes have a one-to-one correspondence.
[0033] Based on the first standard data set, multiple standard data sets are obtained sequentially, and multiple target knowledge graph sets are constructed based on the multiple standard data sets.
[0034] In one embodiment, it also includes:
[0035] Knowledge extraction is performed sequentially on multiple standard datasets to obtain multiple knowledge extraction datasets, wherein the knowledge extraction data includes target entities, connection relationships, and attribute information;
[0036] Relationships between target entities are represented based on multiple knowledge extraction datasets, and the target entities and their relationships are presented in a graphical manner, generating multiple target knowledge graph datasets.
[0037] Data collection results from multiple similar equipment products are sequentially processed through data cleaning, standardization, and partitioning to obtain multiple standard datasets. Knowledge extraction is then performed on these datasets, whereby the extracted data includes target entities, connections, and attribute information. Target entities refer to distinguishable and independently existing entities, such as equipment users or equipment maintenance times. Different entities have different connections. The attribute information describes the characteristics or specific values of the target entities, resulting in multiple knowledge extraction datasets.
[0038] Then, based on multiple knowledge extraction data sets, the relationship between target entities is represented using a bottom-up approach to knowledge graphs. The relationship between target entities is then connected, and the target entities and their relationships are displayed in a graph-like manner, generating multiple target knowledge graph sets. Each target knowledge graph has a one-to-one correspondence with the device usage time node.
[0039] By constructing a target knowledge graph, multiple potential connections can be uncovered when a target device fails, further improving the accuracy of the analysis of the causes of device failures.
[0040] Based on the device usage time nodes, multiple target knowledge graph sets are traversed and compared to obtain a set of graph similar features, wherein the graph similar features correspond to the device usage time nodes.
[0041] In one embodiment, it also includes:
[0042] Randomly select a first target knowledge graph and a second target knowledge graph from multiple sets of target knowledge graphs;
[0043] The first target knowledge graph and the second target knowledge graph are subjected to feature refinement analysis to obtain a first image feature set and a second image feature set;
[0044] Perform feature similarity analysis on the first image feature set and the second image feature set to generate a first atlas similar feature set;
[0045] A set of similar features of the map is obtained based on the first set of similar features of the map.
[0046] Multiple target knowledge graph sets are traversed and compared according to the device usage time nodes. First, a first target knowledge graph and a second target knowledge graph are randomly selected from the multiple target knowledge graph sets, wherein the first target knowledge graph and the second target knowledge graph are different.
[0047] Then, feature refinement analysis is performed on the first target knowledge graph and the second target knowledge graph. The feature refinement analysis refers to extracting the connection relationship and the connected target entity corresponding to each target entity, and taking each target entity, the connected target entity and the connection relationship as an image feature to obtain the first image feature set and the second image feature set.
[0048] Further feature similarity analysis is performed on the first image feature set and the second image feature set. The feature similarity analysis method can be calculated using the existing Manhattan distance similarity calculation method. First, an image feature is randomly selected from the first image feature set as the first feature. Then, the image feature corresponding to the first feature is extracted from the second image feature set as the second feature. Then, the first feature and the second feature are processed into feature vectors, and the Manhattan distance between the two feature vectors is calculated. The smaller the Manhattan distance between the two feature vectors, the higher the similarity between them; the larger the Manhattan distance, the lower the similarity between them.
[0049] A preset feature Manhattan distance threshold is set, which can be set by those skilled in the art based on actual conditions. Similar features in the first image feature set and the second image feature set that are less than the feature Manhattan distance threshold are extracted to generate a first map similar feature set.
[0050] In one embodiment, it also includes:
[0051] Based on the first set of similar features in the map, multiple sets of similar features in the map are obtained sequentially;
[0052] Obtain a similarity feature extraction threshold, and mark the spectral similar features that satisfy the similarity feature extraction threshold in the multiple sets of spectral similar features as high-frequency spectral similar features to obtain multiple high-frequency spectral similar features;
[0053] The spectrum similarity feature set is constructed based on multiple high-frequency spectrum similarity features.
[0054] Two target knowledge graphs are randomly selected from multiple target knowledge graph sets in sequence for image feature similarity analysis to obtain multiple graph similarity feature sets.
[0055] A similarity feature extraction threshold is obtained, which is used to determine the frequency of occurrence of the same feature. This threshold can be set based on the actual amount of feature data; for example, it could be set to 20 occurrences of the same feature. Based on this threshold, the frequency of occurrence of similar features in multiple spectral similarity feature sets is determined. When the frequency of occurrence of a similar feature in a spectral similarity feature set exceeds the threshold, the similar feature is marked as a high-frequency spectral similarity feature, resulting in multiple high-frequency spectral similar features. Then, a spectral similarity feature set is constructed based on these high-frequency spectral similar features, where the spectral similarity features correspond to device usage time points.
[0056] By comparing similar features of multiple target knowledge graph sets and extracting the graph similar features that appear frequently in the comparison results, the accuracy of obtaining graph similar feature sets can be improved, thereby improving the accuracy of abnormal status monitoring of target equipment.
[0057] Construct a twin model of the target device, and embed the graph similarity features in the graph similarity feature set into the target device twin model according to the device usage time nodes;
[0058] The manufacturing and control parameters of the target equipment are obtained. Based on digital twin technology, these parameters are input into digital twin modeling software to model and simulate the target equipment, generating a target equipment twin model. Commonly used digital twin simulation software includes ANSYS Twin Builder, Siemens Simcenter Amesim, etc. Those skilled in the art can select the appropriate simulation software based on the actual situation. Then, the similarity features from the set of similar features are embedded into the target equipment twin model according to the equipment's usage time nodes, generating the required target equipment twin model. Constructing a target equipment twin model based on digital twin technology provides support for real-time monitoring and fault prediction of the target equipment's operation process, and improves the accuracy of target equipment operation status monitoring.
[0059] Based on the target device twin model, fault warnings are provided during the use of the target device.
[0060] In one embodiment, it also includes:
[0061] During the use of the target device, real-time device data is acquired based on the device usage time points;
[0062] The data collected by the real-time device is compared with the set of similar features of the same device usage time nodes in the twin model of the target device.
[0063] During the use of the target equipment, equipment operation data is collected according to the equipment usage time nodes. The equipment operation data includes equipment manufacturing parameters, equipment control parameters, equipment usage environment parameters, and personnel information to obtain real-time equipment data. The real-time equipment data is marked with the equipment usage time node identifier.
[0064] Then, the real-time device data is input into the target device twin model, and the data is compared with the set of similar features of the same device usage time points embedded in the target device twin model.
[0065] When the data collected by the real-time device meets the set of similar features in the graph, a device fault warning signal is generated, and the device is processed according to the device fault warning signal.
[0066] In one embodiment, it also includes:
[0067] Based on the real-time device data, acquire data from multiple devices at previous device usage time points;
[0068] Based on data collected from multiple devices, the target device is simulated and operated using a twin model of the target device to obtain the device simulation results at multiple device usage time points.
[0069] When the simulation results of the equipment meet the preset fault warning threshold, the equipment fault warning signal is generated.
[0070] A feature similarity comparison threshold is set. This threshold can be set by those skilled in the art based on actual conditions; for example, a similarity threshold of 95% can be set. The similarity between the real-time device-collected data and the set of similar features in the map is calculated. The similarity calculation method can be performed using the Manhattan distance similarity calculation method described above, which will not be elaborated here. The similarity calculation result between the real-time device-collected data and the set of similar features in the map is obtained. Then, the similarity calculation result is judged according to the feature similarity comparison threshold. When the similarity calculation result is greater than the feature similarity comparison threshold, multiple device-collected data from previous device usage time points of the target device are obtained.
[0071] Data collected from multiple devices is input into the target device's twin model to simulate the device's operation, obtaining simulation results. A preset fault warning threshold is obtained, which characterizes the features of operational faults in the target device, such as substandard speed or severe abnormal noise. When the simulation results meet the preset fault warning threshold, it indicates a potential risk of device failure. At this point, a device fault warning signal is generated and sent to the maintenance supervisor of the target device, who then promptly detects the fault.
[0072] By constructing a twin model of the target equipment and simulating its operation, it is possible to accurately predict the potential faults that may occur in the target equipment, thereby promptly identifying potential faults and improving the operational stability of the target equipment.
[0073] The above method solves the technical problem of low accuracy and timeliness of existing industrial equipment fault early warning methods. It can improve the accuracy and efficiency of industrial equipment fault early warning, thereby timely detection and handling of potential faults and avoiding serious equipment failures that could affect production.
[0074] In one embodiment, such as Figure 3 The diagram illustrates an industrial knowledge mining and modeling system based on a knowledge graph, comprising:
[0075] A pre-built knowledge graph information acquisition module is used to acquire basic information of the pre-built knowledge graph, wherein the basic information includes graph application scenarios, graph application goals, and graph construction width.
[0076] The data acquisition information determination module is used to determine the data acquisition type and data acquisition quantity based on the basic information. The data acquisition type includes equipment manufacturing parameters, equipment control parameters, equipment operating environment parameters, and personnel information.
[0077] The target knowledge graph set construction module is used to collect relevant data on multiple similar device products of the target device based on the data collection type and the data collection volume, and construct multiple target knowledge graph sets according to the data collection results, wherein the target knowledge graph has a corresponding relationship with the device usage time node;
[0078] The graph similarity feature set acquisition module is used to traverse and compare multiple target knowledge graph sets based on the device usage time node to obtain a graph similarity feature set, wherein the graph similarity features have a corresponding relationship with the device usage time node;
[0079] A target device twin model construction module is used to construct a target device twin model and embed the graph similarity features in the graph similarity feature set into the target device twin model according to the device usage time node;
[0080] A target equipment fault early warning module is used to provide fault early warnings for the target equipment during its use based on the target equipment twin model.
[0081] In one embodiment, the system further includes:
[0082] The first data acquisition result acquisition module is used to randomly select a device product as the first device product from multiple devices of the same type, and acquire the first data acquisition result of the first device product.
[0083] A data cleaning and processing module is used to perform data cleaning and processing on the first data acquisition result to obtain a first data cleaning and processing result. The data cleaning and processing steps include missing value processing, outlier processing, data deduplication, and data format conversion.
[0084] The first standard data set acquisition module is used to further standardize the first data cleaning and processing results, and divide the standard data processing results according to the equipment usage time nodes to obtain the first standard data set, wherein the first standard data and the equipment usage time nodes have a corresponding relationship.
[0085] The target knowledge graph set construction module is used to sequentially obtain multiple standard data sets based on the first standard data set, and construct multiple target knowledge graph sets based on the multiple standard data sets.
[0086] In one embodiment, the system further includes:
[0087] The knowledge extraction module is used to sequentially extract knowledge from multiple standard data sets to obtain multiple knowledge extraction data sets, wherein the knowledge extraction data includes target entities, connection relationships, and attribute information.
[0088] The target knowledge graph set generation module is used to represent the relationship between target entities based on multiple knowledge extraction data sets, and to present the target entities and their relationships in a graphical manner, thereby generating multiple target knowledge graph sets.
[0089] In one embodiment, the system further includes:
[0090] A target knowledge graph selection module is used to randomly select a first target knowledge graph and a second target knowledge graph from multiple sets of target knowledge graphs.
[0091] An image feature set acquisition module is used to perform feature refinement analysis on the first target knowledge graph and the second target knowledge graph to obtain a first image feature set and a second image feature set.
[0092] A feature similarity analysis module is used to perform feature similarity analysis on the first image feature set and the second image feature set to generate a first atlas similar feature set;
[0093] The graph similarity feature set acquisition module is used to obtain a graph similarity feature set based on the first graph similarity feature set.
[0094] In one embodiment, the system further includes:
[0095] The map similarity feature set acquisition module is used to sequentially obtain multiple map similarity feature sets based on the first map similarity feature set;
[0096] A high-frequency spectral similarity feature acquisition module is used to obtain a similarity feature extraction threshold, and to mark spectral similarity features that meet the similarity feature extraction threshold in a plurality of spectral similarity feature sets as high-frequency spectral similarity features, thereby obtaining a plurality of high-frequency spectral similarity features;
[0097] The spectral similarity feature set construction module is used to construct the spectral similarity feature set based on multiple high-frequency spectral similarity features.
[0098] In one embodiment, the system further includes:
[0099] A real-time device data acquisition module is used to acquire real-time device data of the target device based on the device usage time nodes during the use of the target device.
[0100] The traversal comparison module is used to traverse and compare the real-time device data with the set of similar features of the same device usage time nodes in the twin model of the target device.
[0101] The equipment fault warning signal generation module is used to generate an equipment fault warning signal when the real-time equipment collected data meets the map similarity feature set, and to process the equipment according to the equipment fault warning signal.
[0102] In one embodiment, the system further includes:
[0103] The device data acquisition module is used to acquire multiple device data points from previous device usage time points based on the real-time device data.
[0104] The device simulation operation result acquisition module is used to acquire device simulation operation results at multiple device usage time points by simulating the operation of the target device through the target device twin model based on data collected from multiple devices.
[0105] The equipment fault warning signal generation module is used to generate the equipment fault warning signal when the equipment simulation operation result meets the preset fault warning threshold.
[0106] In summary, compared with the prior art, the embodiments of this disclosure have the following technical effects:
[0107] (1) By constructing a twin model of industrial equipment based on knowledge mining, and by using the twin model to simulate and predict equipment faults based on the current operating status of the equipment, the accuracy and efficiency of industrial equipment fault early warning can be improved, thereby timely detection and handling of potential faults and avoiding serious equipment failures that could affect production.
[0108] (2) By cleaning the first data collection results, the data processing time can be reduced, the integrity, uniqueness and accuracy of the data collection results can be improved, and the accuracy of the target knowledge graph construction can be improved. By constructing the target knowledge graph, multiple potential connections when the target equipment fails can be discovered, which further improves the accuracy of the equipment failure cause analysis.
[0109] (3) By comparing the similar features of multiple target knowledge graph sets and extracting the graph similar features that appear frequently in the comparison results, the accuracy of obtaining the graph similar feature set can be improved, thereby improving the accuracy of monitoring the abnormal status of the target equipment.
[0110] (4) By constructing a twin model of the target equipment and simulating the operation of the target equipment, the potential faults of the target equipment can be accurately predicted, thereby timely discovering the potential faults of the target equipment and improving the operational stability of the target equipment.
[0111] The embodiments described above are merely illustrative of several implementations of this disclosure and should not be construed as limiting the scope of the invention. Therefore, those skilled in the art can make various types of substitutions, modifications, and alterations without departing from the scope of the concept as defined by the appended claims, and all such substitutions, modifications, and alterations fall within the protection scope of this disclosure.
Claims
1. A method for industrial knowledge mining and modeling based on knowledge graphs, characterized in that, The method includes: Obtain basic information about the pre-built knowledge graph, including graph application scenarios, graph application goals, and graph construction width; The data acquisition type and data acquisition volume are determined based on the basic information. The data acquisition type includes equipment manufacturing parameters, equipment control parameters, equipment operating environment parameters, and personnel information. Based on the data collection type and the data collection volume, relevant data are collected from multiple similar equipment products of the target device, and multiple target knowledge graph sets are constructed based on the data collection results, wherein the target knowledge graphs correspond to the device usage time nodes; Based on the device usage time nodes, multiple target knowledge graph sets are traversed and compared to obtain a set of graph similar features, wherein the graph similar features correspond to the device usage time nodes. Construct a twin model of the target device, and embed the graph similarity features in the graph similarity feature set into the target device twin model according to the device usage time nodes; Based on the target equipment twin model, fault warnings are provided during the use of the target equipment; The step of traversing and comparing multiple target knowledge graph sets based on the device usage time nodes to obtain a set of graph similar features also includes: Randomly select a first target knowledge graph and a second target knowledge graph from multiple sets of target knowledge graphs; The first target knowledge graph and the second target knowledge graph are subjected to feature refinement analysis to obtain a first image feature set and a second image feature set; Perform feature similarity analysis on the first image feature set and the second image feature set to generate a first atlas similar feature set; The map similarity feature set obtained based on the first map similarity feature set also includes: Based on the first set of similar features in the map, multiple sets of similar features in the map are obtained sequentially; Obtain a similarity feature extraction threshold, and mark the spectral similar features that satisfy the similarity feature extraction threshold in the multiple sets of spectral similar features as high-frequency spectral similar features to obtain multiple high-frequency spectral similar features; The spectrum similarity feature set is constructed based on multiple high-frequency spectrum similarity features.
2. The method as described in claim 1, characterized in that, The process of constructing multiple target knowledge graph sets based on data collection results, previously included: Randomly select one device product from multiple similar device products as the first device product, and obtain the first data acquisition result of the first device product; The first data acquisition result is subjected to data cleaning processing to obtain the first data cleaning processing result, wherein the data cleaning processing steps include missing value processing, outlier processing, data deduplication, and data format conversion; Further, the first data cleaning and processing results are subjected to data standardization processing, and the standard data processing results are divided according to the equipment usage time nodes to obtain the first standard data set, wherein the first standard data and the equipment usage time nodes have a corresponding relationship; Based on the first standard data set, multiple standard data sets are obtained sequentially, and multiple target knowledge graph sets are constructed based on the multiple standard data sets.
3. The method as described in claim 2, characterized in that, The method of constructing multiple target knowledge graph sets based on multiple standard data sets also includes: Knowledge extraction is performed sequentially on multiple standard datasets to obtain multiple knowledge extraction datasets, wherein the knowledge extraction data includes target entities, connection relationships, and attribute information; Relationships between target entities are represented based on multiple knowledge extraction datasets, and the target entities and their relationships are presented in a graphical manner, generating multiple target knowledge graph datasets.
4. The method as described in claim 1, characterized in that, The method of providing fault warnings based on the target device twin model during the use of the target device further includes: During the use of the target device, real-time device data is acquired based on the device usage time points; The data collected by the real-time device is compared with the set of similar features of the same device usage time nodes in the twin model of the target device. When the data collected by the real-time device meets the set of similar features in the graph, a device fault warning signal is generated, and the device is processed according to the device fault warning signal.
5. The method as described in claim 4, characterized in that, The step of generating a device fault warning signal when the real-time device collects data that satisfies the set of similar features in the graph also includes: Based on the real-time device data, acquire data from multiple devices at previous device usage time points; Based on data collected from multiple devices, the target device is simulated and operated using a twin model of the target device to obtain the device simulation results at multiple device usage time points. When the simulation results of the equipment meet the preset fault warning threshold, the equipment fault warning signal is generated.
6. A knowledge graph-based industrial knowledge mining and modeling system, characterized in that, The system is used to perform the steps of any one of the knowledge graph-based industrial knowledge mining and modeling methods described in claims 1-5, wherein the system comprises: A pre-built knowledge graph information acquisition module is used to acquire basic information of the pre-built knowledge graph, wherein the basic information includes graph application scenarios, graph application goals, and graph construction width. The data acquisition information determination module is used to determine the data acquisition type and data acquisition quantity based on the basic information. The data acquisition type includes equipment manufacturing parameters, equipment control parameters, equipment operating environment parameters, and personnel information. The target knowledge graph set construction module is used to collect relevant data on multiple similar device products of the target device based on the data collection type and the data collection volume, and construct multiple target knowledge graph sets according to the data collection results, wherein the target knowledge graph has a corresponding relationship with the device usage time node; The graph similarity feature set acquisition module is used to traverse and compare multiple target knowledge graph sets based on the device usage time node to obtain a graph similarity feature set, wherein the graph similarity features have a corresponding relationship with the device usage time node; A target device twin model construction module is used to construct a target device twin model and embed the graph similarity features in the graph similarity feature set into the target device twin model according to the device usage time node; A target equipment fault early warning module is used to provide fault early warning for the use of the target equipment based on the target equipment twin model. The system also includes: A target knowledge graph selection module is used to randomly select a first target knowledge graph and a second target knowledge graph from multiple sets of target knowledge graphs. An image feature set acquisition module is used to perform feature refinement analysis on the first target knowledge graph and the second target knowledge graph to obtain a first image feature set and a second image feature set. A feature similarity analysis module is used to perform feature similarity analysis on the first image feature set and the second image feature set to generate a first atlas similar feature set; The map similarity feature set acquisition module is used to obtain a map similarity feature set based on the first map similarity feature set. The map similarity feature set acquisition module is used to sequentially obtain multiple map similarity feature sets based on the first map similarity feature set; A high-frequency spectral similarity feature acquisition module is used to obtain a similarity feature extraction threshold, and to mark spectral similarity features that meet the similarity feature extraction threshold in a plurality of spectral similarity feature sets as high-frequency spectral similarity features, thereby obtaining a plurality of high-frequency spectral similarity features; The spectral similarity feature set construction module is used to construct the spectral similarity feature set based on multiple high-frequency spectral similarity features.