An Industry Knowledge Structured Modeling Method Based on a Semantically Associated Knowledge Graph

Abstract

The present invention belongs to the technical field of knowledge graphs, and discloses an industry knowledge structured modeling method for a knowledge graph based on semantic association; collecting industry production data; constructing a production knowledge graph according to the industry production data; obtaining personalized demand content; extracting features from the personalized demand content to obtain actual demand feature data; inputting the actual demand feature data into a demand evaluation model to obtain a demand evaluation score, and analyzing and processing the demand evaluation score to obtain a demand rationality level; inputting the actual demand feature data and the production knowledge graph into a parameter construction model to obtain a corresponding production parameter set; inputting the demand rationality level, the production parameter set and the production knowledge graph into a demand construction model to obtain predicted demand feature data; dynamically evaluating the matching degree between customer demands and system predictions, and generating a production parameter set that meets the personalized demands of customers, so as to improve production accuracy.

Description

Technical Field

[0001] The present invention relates to the field of knowledge graph technology, and more specifically, to a method for structured modeling of industry knowledge based on a knowledge graph with semantic associations. Background Art

[0002] A variety of intelligent devices are widely used in enterprise production activities. These devices can generate massive multi-source data in real time. These data cover different types, dimensions and structures, involving a large spatial and temporal span of production, and are mainly used to improve production reliability, repeatability and product quality. Business process data modeling aims to accurately identify bottlenecks and waste links in the production process through deep integration and analysis of multi-source data, and propose corresponding improvement strategies based on data-driven, so as to achieve efficiency and flexibility of business processes. By constructing multi-source data into a business process model, it can effectively support the quantitative analysis of lean production processes, optimize process structures, and improve production efficiency and overall benefits. Therefore, how to build a unified business process data model in a distributed information system environment and realize dynamic analysis and optimization of complex business processes through the model is a key technical problem that needs to be solved in this field.

[0003] In order to improve the flexibility and application adaptability of knowledge graphs in complex environments, the current Chinese patent application with publication number CN117992619A proposes an industry knowledge graph modeling method and system based on an attribute graph, the method comprising: creating basic attributes and basic relationships of the industry knowledge graph, wherein all data is stored using an attribute graph database as a storage engine; creating business attributes and business relationships required for industry knowledge applications by combining basic attributes and basic relationships; filling in business relationships between different industry knowledge ontologies to form an ontology relationship graph, combining business attributes and business relationships, and adding business knowledge constraints to form an industry knowledge ontology; creating extensible industry knowledge instances based on the industry knowledge ontology; filling in business relationships between different industry knowledge instances to form an instance relationship graph; forming an industry knowledge graph based on the industry knowledge ontology, the ontology relationship graph, the industry knowledge instance, and the instance relationship graph to realize industry knowledge business.

[0004] Although the above method can meet most scenarios, research and practical application of the above method and existing technology have found that the above method and existing technology have at least the following defects:

[0005] The existing technology shows obvious limitations in dealing with diverse and personalized needs. The production process is difficult to be dynamically adjusted according to customer requirements and cannot achieve a quick response to complex demand scenarios. The setting of production parameters usually relies on manual analysis or adjustment methods based on fixed rules, lacking intelligent optimization capabilities, resulting in the inability to achieve dynamic adjustment and real-time feedback, large fluctuations in product quality, and waste of production materials. Furthermore, it leads to incomplete matching between product parameter settings and customer requirements, affecting the customer experience and market competitiveness.

[0006] In view of this, the present invention proposes an industry knowledge structured modeling method based on a knowledge graph of semantic associations to solve the above problems. Summary of the Invention

[0007] To overcome the above defects of the existing technology and to achieve the above object, the present invention provides the following technical solution: An industry knowledge structured modeling method based on a knowledge graph of semantic associations, comprising:

[0008] Collect industry production data;

[0009] Construct a production knowledge graph according to the industry production data;

[0010] Obtain personalized demand content;

[0011] Extract features from the personalized demand content to obtain actual demand feature data;

[0012] Input the actual demand feature data into a demand evaluation model to obtain a demand evaluation score, and analyze and process the demand evaluation score to obtain a demand rationality level;

[0013] Input the actual demand feature data and the production knowledge graph into a parameter construction model to obtain a corresponding set of production parameters;

[0014] Input the demand rationality level, the set of production parameters, and the production knowledge graph into a demand construction model to obtain predicted demand feature data;

[0015] Extract features from the actual demand feature data and the predicted demand feature data to obtain a demand matching degree;

[0016] Comprehensively analyze the demand matching degree, the set of production parameters, and the production knowledge graph, generate an updated set of production parameters, send an instruction to the intelligent factory, and adjust the production parameters to the updated set of production parameters for production.

[0017] Furthermore, the method for comprehensively analyzing the demand matching degree, the set of production parameters, and the production knowledge graph includes:

[0018] Preset a demand matching degree threshold;

[0019] If the demand matching degree is greater than or equal to the demand matching degree threshold, the actual demand feature data matches the predicted demand feature data, and the production parameter set is used as the updated production parameter set;

[0020] If the demand matching degree is less than the demand matching degree threshold, the actual demand feature data does not match the predicted demand feature data. A demand mismatch report is generated. After the customer reconfirms the demand, the updated actual demand feature data is obtained. The updated actual demand feature data and the production knowledge graph are input into the parameter construction model to obtain the corresponding production parameter set, which is marked as the production parameter set to be changed, and the production parameter set to be changed is used as the updated production parameter set;

[0021] Send an instruction to the intelligent factory to adjust the production parameters to the updated production parameter set and perform production.

[0022] Furthermore, the method for obtaining the demand rationality level includes:

[0023] Preset an evaluation score threshold and , less than ; The demand rationality level includes third-level rationality, second-level rationality, and first-level rationality;

[0024] If the demand evaluation score is greater than or equal to , the demand rationality level is third-level rationality, and the system does not issue a demand irrational instruction;

[0025] If the demand evaluation score is greater than or equal to , and less than , the demand rationality level is second-level rationality, and the system does not issue a demand irrational instruction;

[0026] If the demand evaluation score is less than , the demand rationality level is first-level rationality, and the system issues a demand irrational instruction.

[0027] Furthermore, the method for obtaining the demand matching degree includes:

[0028] Subtract the predicted product type in the predicted demand feature data from the expected product type in the actual demand feature data to obtain the product type difference; subtract the predicted cost in the predicted demand feature data from the expected cost in the actual demand feature data to obtain the cost difference; subtract the predicted qualified rate in the predicted demand feature data from the expected qualified rate in the actual demand feature data to obtain the qualified rate difference; subtract the predicted accuracy in the predicted demand feature data from the expected accuracy in the actual demand feature data to obtain the accuracy difference; subtract the predicted delivery cycle in the predicted demand feature data from the expected delivery cycle in the actual demand feature data to obtain the delivery cycle difference;

[0029] Perform a weighted operation on the product type difference, cost difference, qualified rate difference, accuracy difference, and delivery cycle difference to obtain the first result of demand matching degree; sum the squares of the weight coefficients corresponding to the product type difference, cost difference, qualified rate difference, accuracy difference, and delivery cycle difference to obtain the second result of demand matching degree;

[0030] Divide the first result of demand matching degree by the second result of demand matching degree to obtain the third result of demand matching degree, and perform a square root operation on the third result of demand matching degree to obtain the demand matching degree.

[0031] Furthermore, the method for obtaining the actual demand feature data includes:

[0032] Extract the expected product type from the personalized demand content and create a product type key-value pair, where the value of the product type key-value pair is the expected product type in the personalized demand content; extract the expected cost from the personalized demand content and create a cost key-value pair, where the value of the cost key-value pair is the expected cost in the personalized demand content;

[0033] Extract the expected qualified rate from the personalized demand content and create a qualified rate key-value pair, where the value of the qualified rate key-value pair is the expected qualified rate in the personalized demand content; extract the expected accuracy from the personalized demand content and create an accuracy key-value pair, where the value of the accuracy key-value pair is the expected accuracy in the personalized demand content; extract the expected delivery cycle from the personalized demand content and create a delivery cycle key-value pair, where the value of the delivery cycle key-value pair is the expected delivery cycle in the personalized demand content;

[0034] Construct the product type key-value pair, cost key-value pair, qualified rate key-value pair, accuracy key-value pair, and delivery cycle key-value pair into the actual demand feature data.

[0035] Furthermore, the method for constructing a production knowledge graph based on the industry production data includes:

[0036] Convert the temperature, pressure, processing speed, cutting depth, and coating thickness in the production process variables into corresponding knowledge nodes and add them to the first knowledge node set; convert the equipment load, current, voltage, and vibration frequency in the equipment operation data into corresponding knowledge nodes and add them to the second knowledge node set; convert the pass rate, precision, strength, and sealing performance in the production quality indicators into corresponding knowledge nodes and add them to the third knowledge node set; convert the composition, specification, ductility, and density in the raw material characteristics into corresponding knowledge nodes and add them to the fourth knowledge node set;

[0037] Use natural language processing technology to process industry production documents to obtain the relationships between knowledge nodes; the relationships between the knowledge nodes include influence, determination, limitation, and association;

[0038] Construct a knowledge graph from the first knowledge node set, the second knowledge node set, the third knowledge node set, the fourth knowledge node set, and the relationships between the knowledge nodes to obtain a production knowledge graph.

[0039] Further, the training method of the demand assessment model includes:

[0040] Pre-collect a demand assessment data set, the demand assessment data set includes G groups of demand assessment data, and the corresponding demand assessment scores for the G groups of demand assessment data, G is a positive integer greater than 0, and the demand assessment data includes actual demand feature data; divide the demand assessment data set into a training set and a test set, use the demand assessment data in the training set as the input of the demand assessment model, and use the demand assessment scores in the training set as the output of the demand assessment model, and use minimizing the sum of the prediction accuracies of all predicted demand assessment scores as the training goal; stop training until the sum of the prediction accuracies reaches convergence; the demand assessment model is a Naive Bayes model or a Support Vector Machine model.

[0041] Further, the training method of the demand construction model includes:

[0042] Pre-collect a demand construction data set, the demand construction data set includes K groups of demand construction data, and the corresponding predicted demand feature data for the K groups of demand construction data, K is a positive integer greater than 0, and the demand construction data includes demand rationality level, production parameter set, and production knowledge graph; divide the demand construction data set into a training set and a test set, use the demand construction data in the training set as the input of the demand construction model, and use the predicted demand feature data in the training set as the output of the demand construction model, and use minimizing the sum of the prediction accuracies of all predicted predicted demand feature data as the training goal; stop training until the sum of the prediction accuracies reaches convergence; the demand construction model is a Random Forest model or a Recurrent Neural Network model.

[0043] Further, the training method of the parameter construction model includes:

[0044] Pre-collect a parameter construction data set, where the parameter construction data set includes H groups of parameter construction data and the corresponding production parameter sets for the H groups of parameter construction data. H is a positive integer greater than 0. The parameter construction data includes actual demand feature data and a production knowledge graph. Divide the parameter construction data set into a training set and a test set. Use the parameter construction data in the training set as the input of the parameter construction model, and use the production parameter sets in the training set as the output of the parameter construction model. Take minimizing the sum of the prediction accuracies of all predicted production parameter sets as the training objective. Stop training until the sum of the prediction accuracies reaches convergence. The parameter construction model is a CaffeNet model.

[0045] Further, the industrial production data includes production process variables, equipment operation status, production quality indicators, and raw material characteristics. The production process variables include temperature, pressure, processing speed, cutting depth, and coating thickness. The equipment operation data includes equipment load, current, voltage, and vibration frequency. The production quality indicators include pass rate, precision, strength, and sealing performance. The raw material characteristics include composition, specification, ductility, and density. The personalized demand content includes expected product type, expected cost, expected pass rate, expected precision, and expected delivery cycle.

[0046] The technical effects and advantages of an industrial knowledge structured modeling method based on a knowledge graph with semantic association in the present invention:

[0047] By introducing knowledge graph technology, complex production knowledge, process parameters, and actual data in an intelligent factory are associated and modeled, so as to achieve the goal of flexible and efficient production through personalized customization requirements. Through the dynamic evaluation of the demand matching degree, quickly judge the matching degree between customer needs and system predictions. Combine the demand rationality level, production parameter sets, and production knowledge graph for comprehensive analysis to generate the optimal production parameter sets that meet the personalized needs of customers, avoid production errors caused by deviations, and improve production accuracy and customer satisfaction. By adaptively optimizing production parameters, reduce manual intervention, shorten the adjustment time, and ensure that the rule setting and algorithm implementation of each step have sufficient flexibility and accuracy. Meet diverse and personalized production requirements and achieve flexible manufacturing. By real-time adjusting and optimizing the production process, reduce resource waste and unnecessary cost expenditures.

[0048] This solution realizes a closed-loop control from customer needs to production parameter optimization by combining knowledge graph technology, demand evaluation models, and parameter optimization, comprehensively improving the production efficiency, flexibility, and customer satisfaction of intelligent factories, and has broad application value and technological leadership in the field of intelligent manufacturing. Description of the Drawings

[0049] Figure 1 Schematic diagram of an industrial knowledge structured modeling system based on a semantic - related knowledge graph for Embodiment 1 of the present invention;

[0050] Figure 2 Flowchart of a method for industrial knowledge structured modeling based on a semantic - related knowledge graph for Embodiment 2 of the present invention;

[0051] Figure 3 Flowchart of a method for comprehensively analyzing demand matching degree, production parameter set, and production knowledge graph;

[0052] Figure 4 Flowchart of a method for obtaining the rationality level of demand;

[0053] Figure 5 Schematic diagram of a production knowledge graph. Detailed implementation manners

[0054] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0055] Embodiment 1

[0056] Please refer to Figure 1 As shown, the industrial knowledge structured modeling system based on a semantic - related knowledge graph in this embodiment includes a first collection module, a first processing module, a second collection module, a second processing module, a demand evaluation module, a parameter construction module, a demand construction module, a demand matching module, and a parameter determination module. Each module is connected by wire and / or wirelessly to achieve data transmission.

[0057] The first collection module is used to collect industrial production data; the industrial production data includes production process variables, equipment operation status, production quality indicators, and raw material characteristics; the production process variables include temperature, pressure, processing speed, cutting depth, and coating thickness; the equipment operation data includes equipment load, current, voltage, and vibration frequency; the production quality indicators include qualification rate, precision, strength, and sealing performance; the raw material characteristics include composition, specification, ductility, and density.

[0058] The industrial production data is obtained through an intelligent factory or the enterprise's background data management system.

[0059] The first processing module is used to construct a production knowledge graph based on the industrial production data.

[0060] A method for constructing a production knowledge graph based on the industrial production data includes:

[0061] Convert the temperature, pressure, processing speed, cutting depth, and coating thickness in the production process variables into corresponding knowledge nodes and add them to the first knowledge node set; convert the equipment load, current, voltage, and vibration frequency in the equipment operation data into corresponding knowledge nodes and add them to the second knowledge node set; convert the pass rate, precision, strength, and sealing performance in the production quality indicators into corresponding knowledge nodes and add them to the third knowledge node set; convert the composition, specifications, ductility, and density in the raw material characteristics into corresponding knowledge nodes and add them to the fourth knowledge node set;

[0062] For example, convert the temperature into a temperature knowledge node, the pressure into a pressure knowledge node, the processing speed into a processing speed knowledge node, the cutting depth into a cutting depth knowledge node, and the coating thickness into a coating thickness knowledge node.

[0063] Use natural language processing technology to process the industrial production documents to obtain the relationships between the knowledge nodes; the relationships between the knowledge nodes include influence, determination, limitation, and association;

[0064] Construct the first knowledge node set, the second knowledge node set, the third knowledge node set, the fourth knowledge node set, and the relationships between the knowledge nodes into a knowledge graph, that is, obtain the production knowledge graph.

[0065] For example, the heating temperature affects the pass rate, and the relationship between the heating temperature and the pass rate can be obtained; the process variables affect the quality indicators, the material characteristics affect the quality indicators, the equipment status affects the process variables, and the equipment status is associated with the process variables.

[0066] It should be noted that Figure 5 is a schematic diagram of the production knowledge graph. The production documents refer to the records, text descriptions, or technical documents related to the production process. These contents exist in the form of natural language, such as production logs, equipment manuals, operation manuals, process specifications, etc. These documents contain a large amount of information about production processes, equipment operation, quality control, and raw material characteristics.

[0067] The second acquisition module is used to obtain personalized demand content; the personalized demand content includes the expected product type, expected cost, expected pass rate, expected precision, and expected delivery cycle.

[0068] The second processing module is used to extract features from the personalized demand content to obtain actual demand feature data.

[0069] The method for obtaining the actual demand feature data includes:

[0070] Extract the expected product type from the personalized demand content to create a product type key-value pair. The key of the product type key-value pair is "expected product type", and the value of the product type key-value pair is the expected product type in the personalized demand content. Extract the expected cost from the personalized demand content to create a cost key-value pair. The key of the cost key-value pair is "expected cost", and the value of the cost key-value pair is the expected cost in the personalized demand content.

[0071] Extract the expected pass rate from the personalized demand content to create a pass rate key-value pair. The key of the pass rate key-value pair is "expected pass rate", and the value of the pass rate key-value pair is the expected pass rate in the personalized demand content. Extract the expected precision from the personalized demand content to create a precision key-value pair. The key of the precision key-value pair is "expected precision", and the value of the precision key-value pair is the expected precision in the personalized demand content. Extract the expected delivery cycle from the personalized demand content to create a delivery cycle key-value pair. The key of the delivery cycle key-value pair is "expected delivery cycle", and the value of the delivery cycle key-value pair is the expected delivery cycle in the personalized demand content.

[0072] Construct the product type key-value pair, cost key-value pair, pass rate key-value pair, precision key-value pair, and delivery cycle key-value pair into actual demand feature data.

[0073] It should be noted that each key-value pair corresponds to a specific demand feature of the customer, enabling the demand features to be quantified into specific numerical values or categories, facilitating multi-dimensional comparison and analysis, being able to well express the personalized needs of the customer, thus achieving the effect of personalized modeling and meeting the needs of different customers. The demand features in the form of key-value pairs can be directly used as feature vectors and input into the machine learning model for demand prediction, classification, or optimization.

[0074] An example of the actual demand feature data is as follows:

[0075] "actual demand feature data":{

[0076] / / Product type key-value pair

[0077] {"expected product type": "threaded steel pipe"},

[0078] / / Cost key-value pair

[0079] {"expected cost": "5 million"},

[0080] / / Pass rate key-value pair

[0081] {"expected pass rate": "95%"},

[0082] / / Precision key-value pair

[0083] {"Expected accuracy": "0.9"},

[0084] / / Delivery cycle key-value pair

[0085] {"Expected delivery cycle": "From January 10, 2024 to February 14, 2024"}

[0086] };

[0087] A requirement evaluation module, which is used to input actual requirement feature data into a requirement evaluation model to obtain a requirement evaluation score, and analyze and process the requirement evaluation score to obtain a requirement rationality level.

[0088] The training method of the requirement evaluation model includes:

[0089] Pre-collect a requirement evaluation data set, the requirement evaluation data set includes G groups of requirement evaluation data, and the requirement evaluation scores corresponding to the G groups of requirement evaluation data, G is a positive integer greater than 0, and the requirement evaluation data includes actual requirement feature data; the requirement evaluation score is evaluated by multiple technicians in the field according to the actual requirement feature data, and corresponding scoring is carried out. After removing the maximum value and the minimum value, the average value is taken to obtain the requirement evaluation score; divide the requirement evaluation data set into a training set and a test set, use the requirement evaluation data in the training set as the input of the requirement evaluation model, and use the requirement evaluation scores in the training set as the output of the requirement evaluation model, and use minimizing the sum of the prediction accuracies of all predicted requirement evaluation scores as the training goal; stop training until the sum of the prediction accuracies reaches convergence; the requirement evaluation model is a Naive Bayes model or a Support Vector Machine model.

[0090] As Figure 4 shown, the method for obtaining the requirement rationality level includes:

[0091] Preset an evaluation score threshold and , less than ; the requirement rationality level includes third-level rationality, second-level rationality and first-level rationality. The higher the rationality level, the higher the requirement rationality;

[0092] If the requirement evaluation score is greater than or equal to , the requirement content is reasonable, the requirement rationality level is third-level rationality, and the system does not issue a requirement unreasonable instruction;

[0093] If the requirement evaluation score is greater than or equal to , and less than , the requirement content is slightly unreasonable, the requirement rationality level is second-level rationality, and the system does not issue a requirement unreasonable instruction;

[0094] If the demand assessment score is less than , the demand content is seriously unreasonable, the demand rationality level is the first-level rationality, and the system issues a demand unreasonable instruction.

[0095] It should be noted that by setting multi-level thresholds, the rationality level of demands is accurately divided, avoiding a one-size-fits-all simple judgment method, and improving the flexibility and applicability of evaluation. When the demand is seriously unreasonable, the system can issue a warning in time and terminate the production process, avoiding resource waste and potential losses, and enhancing the safety and reliability of the system; for slightly unreasonable demands, the system allows continued operation, avoiding excessive interference with production efficiency, and at the same time providing room for optimizing minor problems. The dynamic judgment mechanism based on the demand assessment score provides a scientific decision-making basis for production managers and reduces the impact of subjective judgment on the production process. By clarifying the demand rationality judgment criteria and hierarchical response mechanism, customers can optimize their demands according to the system feedback, so as to obtain a solution that meets expectations faster. This hierarchical judgment mechanism realizes the intelligence and controllability of demand rationality judgment, and provides strong technical support for improving the intelligence level and flexibility of the production system.

[0096] The parameter construction module is used to input the actual demand feature data and the production knowledge graph into the parameter construction model to obtain the corresponding production parameter set.

[0097] The training method of the parameter construction model includes:

[0098] Pre-collect a parameter construction data set, the parameter construction data set includes H groups of parameter construction data, and the production parameter set corresponding to the H groups of parameter construction data, H is a positive integer greater than 0, and the parameter construction data includes actual demand feature data and a production knowledge graph; divide the parameter construction data set into a training set and a test set, use the parameter construction data in the training set as the input of the parameter construction model, and use the production parameter set in the training set as the output of the parameter construction model, and minimize the sum of the prediction accuracies of all predicted production parameter sets as the training goal; stop training until the sum of the prediction accuracies reaches convergence; the parameter construction model is a CaffeNet model.

[0099] It should be noted that the actual demand feature data includes multi-dimensional demand indicators such as product type, cost, qualified rate, precision, and delivery cycle, representing the personalized requirements of customers for products. The production knowledge graph constructs multi-level production factors such as production processes, equipment operation parameters, quality control indicators, and raw material characteristics and the relationships between them through nodes and edges. Through the parameter construction model, the actual demand feature data can be mapped to the corresponding nodes of the production knowledge graph to complete the semantic matching and association of demand features and production factors. The production knowledge graph contains constraint rules between production factors (for example, higher precision requirements may reduce processing speed; shorter delivery cycles may increase equipment operation load). The parameter construction model uses these rules to reason and evaluate the input requirements to determine a reasonable set of production parameters.

[0100] A demand construction module for inputting the demand rationality level, the set of production parameters, and the production knowledge graph into a demand construction model to obtain predicted demand feature data.

[0101] The training method of the demand construction model includes:

[0102] Pre-collect a demand construction data set, where the demand construction data set includes K groups of demand construction data and the predicted demand feature data corresponding to the K groups of demand construction data. K is a positive integer greater than 0. The demand construction data includes the demand rationality level, the set of production parameters, and the production knowledge graph; divide the demand construction data set into a training set and a test set, use the demand construction data in the training set as the input of the demand construction model, and use the predicted demand feature data in the training set as the output of the demand construction model. Take minimizing the sum of the prediction accuracies of all predictions of the predicted demand feature data as the training objective; stop training until the sum of the prediction accuracies reaches convergence; the demand construction model is a random forest model or a recurrent neural network model.

[0103] It should be noted that when the demand rationality level is evaluated as secondary rationality, the customer demand is in a slightly unreasonable range, which may cause a certain deviation between the demand feature data predicted by the system and the actual customer demand, thus affecting the accuracy of production decisions and the ability to meet customer personalized needs. To avoid this problem and improve the matching degree between the demand feature data and the actual customer demand, the present invention proposes an optimization strategy based on comprehensive analysis, which combines the demand rationality level, the set of production parameters, and the production knowledge graph for multi-dimensional correlation reasoning and dynamic optimization.

[0104] By comprehensively considering the rationality level of requirements, the model can identify the reasonable range of customer requirements; combined with the production parameter set, it can evaluate the response level of the current production capacity to the requirements; using the production knowledge graph, it can conduct correlation analysis on the logical relationships and adjustable spaces among various production factors. Based on this, the system uses a constraint optimization algorithm to correct the demand feature data and derives the optimal demand feature data achievable under the current production conditions, ensuring a balance between technical feasibility and demand matching degree.

[0105] A demand matching module, which is used to extract features from the actual demand feature data and the predicted demand feature data to obtain the demand matching degree.

[0106] The method for obtaining the demand matching degree includes:

[0107] ;

[0108] Among them, is the demand matching degree, is the predicted product type in the predicted demand feature data, is the expected product type, is the predicted cost in the predicted demand feature data, is the expected cost, is the predicted qualified rate in the predicted demand feature data, is the expected qualified rate, is the predicted accuracy in the predicted demand feature data, is the expected accuracy, is the predicted delivery cycle in the predicted demand feature data, is the expected delivery cycle, 、 、 and are weight coefficients.

[0109] A parameter determination module, which is used to comprehensively analyze the demand matching degree, the production parameter set and the production knowledge graph, generate an updated production parameter set, send an instruction to the intelligent factory, and adjust the production parameters to the updated production parameter set for production.

[0110] As Figure 3 shown, the method for comprehensively analyzing the demand matching degree, the production parameter set and the production knowledge graph includes:

[0111] Preset a demand matching degree threshold;

[0112] If the demand matching degree is greater than or equal to the demand matching degree threshold, the actual demand feature data matches the predicted demand feature data, and the production parameter set is used as the updated production parameter set;

[0113] If the demand matching degree is less than the demand matching degree threshold, the actual demand feature data does not match the predicted demand feature data. A demand mismatch report is generated. After the customer reconfirms the demand, the updated actual demand feature data is obtained. The updated actual demand feature data and the production knowledge graph are input into the parameter construction model to obtain the corresponding production parameter set, which is marked as the production parameter set to be changed. The production parameter set to be changed is used as the updated production parameter set.

[0114] Send an instruction to the intelligent factory to adjust the production parameters to the updated production parameter set and carry out production.

[0115] It should be noted that by setting the matching degree threshold, the matching situation between the actual demand feature data and the predicted demand feature data can be quickly judged, ensuring that production can only enter the production link when the demand is accurate, thereby reducing production errors caused by demand deviation and improving production accuracy. For the situation where the demand matching degree is insufficient, the system generates a demand mismatch report and requires the customer to reconfirm, ensuring that the corrected demand can be closer to the actual expectations of the customer, improving the demand adaptation ability and customer satisfaction. Combining the updated actual demand feature data with the production knowledge graph, a new production parameter set is dynamically generated, and the production plan is optimized through an intelligent adjustment model to ensure that the production process can flexibly adapt to demand changes.

[0116] For the situation where the demand matching degree is sufficient, the existing production parameter set is directly adopted to simplify the process; for the situation where the demand matching degree is insufficient, updated production parameters are quickly generated through an intelligent adjustment mechanism, reducing human intervention and improving the overall production efficiency and response speed. A closed-loop control is realized throughout the whole process from demand assessment to production parameter adjustment and then to final production, ensuring that production operations meet customer needs, and at the same time providing a technical basis for dynamic optimization, laying a solid foundation for intelligent production.

[0117] Embodiment 2

[0118] Please refer to Figure 2 As shown in the figure, this embodiment provides an industry knowledge structured modeling method based on a knowledge graph of semantic association, and further includes:

[0119] Collect industry production data;

[0120] Construct a production knowledge graph according to the industry production data;

[0121] Obtain personalized demand content;

[0122] Extract features from the personalized demand content to obtain actual demand feature data;

[0123] Input the actual demand feature data into the demand assessment model to obtain a demand assessment score, and analyze and process the demand assessment score to obtain the demand rationality level;

[0124] Input the actual demand feature data and the production knowledge graph into the parameter construction model to obtain the corresponding production parameter set;

[0125] Input the demand rationality level, the production parameter set and the production knowledge graph into the demand construction model to obtain the predicted demand feature data;

[0126] Extract features from the actual demand feature data and the predicted demand feature data to obtain the demand matching degree;

[0127] Conduct a comprehensive analysis of the demand matching degree, the production parameter set and the production knowledge graph to generate an updated production parameter set, send an instruction to the intelligent factory to adjust the production parameters to the updated production parameter set and carry out production.

[0128] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

[0129] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for structural modeling of industry knowledge based on a knowledge graph with semantic association, characterized in that: include: Collect industry production data; Build a production knowledge graph based on industry production data; Get personalized demand content; Extract features from personalized demand content to obtain actual demand feature data; Input the actual demand characteristic data into the demand assessment model to obtain the demand assessment score, and analyze and process the demand assessment score to obtain the demand rationality level; Input the actual demand feature data and production knowledge graph into the parameter construction model to obtain the corresponding production parameter set; Input the demand rationality level, production parameter set and production knowledge graph into the demand construction model to obtain the predicted demand feature data; Extract features from actual demand feature data and predicted demand feature data to obtain demand matching degree; Comprehensively analyze the demand matching degree, production parameter set and production knowledge graph, generate an updated production parameter set, send instructions to the smart factory, adjust the production parameters to the updated production parameter set and carry out production; Methods for comprehensive analysis of demand matching, production parameter sets, and production knowledge graphs include: Preset demand matching threshold; If the demand matching degree is greater than or equal to the demand matching degree threshold, the actual demand feature data matches the predicted demand feature data, and the production parameter set is used as the updated production parameter set; If the demand matching degree is less than the demand matching degree threshold, the actual demand feature data does not match the predicted demand feature data, and a demand mismatch report is generated. After the customer reconfirms the demand, the updated actual demand feature data is obtained; the updated actual demand feature data and the production knowledge graph are input into the parameter construction model to obtain the corresponding production parameter set, which is marked as the production parameter set to be changed, and the production parameter set to be changed is used as the updated production parameter set; Send instructions to the smart factory to adjust the production parameters to the updated production parameter set and start production.

2. According to claim 1, a method for structurally modeling industry knowledge based on a knowledge graph of semantic association, characterized in that: The method for obtaining the demand rationality level includes: Preset evaluation score thresholds and , Less than ; The demand rationality levels include third-level rationality, second-level rationality and first-level rationality; If the needs assessment score is greater than or equal to , then the demand rationality level is level 3 rationality, and the system does not issue unreasonable demand instructions; If the needs assessment score is greater than or equal to , and less than , then the demand rationality level is level 2 rationality, and the system does not issue unreasonable demand instructions; If the needs assessment score is less than , then the demand rationality level is level one rationality, and the system issues an unreasonable demand instruction.

3. The method for structurally modeling industry knowledge based on a semantically associated knowledge graph according to claim 2 is characterized in that: The method for obtaining the demand matching degree includes: Subtract the predicted product type in the predicted demand feature data from the expected product type in the actual demand feature data to obtain the product type difference; subtract the predicted cost in the predicted demand feature data from the expected cost in the actual demand feature data to obtain the cost difference; subtract the predicted pass rate in the predicted demand feature data from the expected pass rate in the actual demand feature data to obtain the pass rate difference; subtract the predicted accuracy in the predicted demand feature data from the expected accuracy in the actual demand feature data to obtain the accuracy difference; subtract the predicted delivery cycle in the predicted demand feature data from the expected delivery cycle in the actual demand feature data to obtain the delivery cycle difference; Perform weighted operations on the product type difference, cost difference, pass rate difference, precision difference and delivery cycle difference to obtain the first result of demand matching; sum the squares of the weight coefficients corresponding to the product type difference, cost difference, pass rate difference, precision difference and delivery cycle difference to obtain the second result of demand matching; The first result of the demand matching degree is divided by the second result of the demand matching degree to obtain a third result of the demand matching degree, and the third result of the demand matching degree is squared to obtain the demand matching degree.

4. The method for structurally modeling industry knowledge based on a semantically associated knowledge graph according to claim 3 is characterized in that: The method for acquiring the actual demand feature data includes: Extract the expected product type from the personalized demand content, create a product type key-value pair, the value of the product type key-value pair is the expected product type in the personalized demand content; extract the expected cost from the personalized demand content, create a cost key-value pair, the value of the cost key-value pair is the expected cost in the personalized demand content; Extract the expected pass rate from the personalized demand content, create a pass rate key-value pair, the value of which is the expected pass rate in the personalized demand content; extract the expected accuracy from the personalized demand content, create a accuracy key-value pair, the value of which is the expected accuracy in the personalized demand content; extract the expected delivery cycle from the personalized demand content, create a delivery cycle key-value pair, the value of which is the expected delivery cycle in the personalized demand content; The product type key-value pair, cost key-value pair, pass rate key-value pair, precision key-value pair and delivery cycle key-value pair are constructed into actual demand feature data.

5. According to claim 4, a method for structural modeling of industry knowledge based on a knowledge graph of semantic association, characterized in that: The method for constructing a production knowledge graph based on the industry production data includes: The temperature, pressure, processing speed, cutting depth and coating thickness in the production process variables are converted into corresponding knowledge nodes and added to the first knowledge node set; the equipment load, current, voltage and vibration frequency in the equipment operation data are converted into corresponding knowledge nodes and added to the second knowledge node set; the qualified rate, precision, strength and sealing in the production quality indicators are converted into corresponding knowledge nodes and added to the third knowledge node set; the composition, specification, ductility and density in the raw material characteristics are converted into corresponding knowledge nodes and added to the fourth knowledge node set; Using natural language processing technology to process industry production documents to obtain relationships between knowledge nodes; the relationships between knowledge nodes include influence, determination, restriction and association; The first knowledge node set, the second knowledge node set, the third knowledge node set, the fourth knowledge node set and the relationships between the knowledge nodes are constructed into a knowledge graph to obtain a production knowledge graph.

6. The method for structurally modeling industry knowledge based on a semantically associated knowledge graph according to claim 5, characterized in that: The training method of the demand assessment model includes: A demand assessment data set is collected in advance, wherein the demand assessment data set includes G groups of demand assessment data and demand assessment scores corresponding to the G groups of demand assessment data, where G is a positive integer greater than 0, and the demand assessment data includes actual demand feature data; the demand assessment data set is divided into a training set and a test set, the demand assessment data in the training set is used as the input of the demand assessment model, and the demand assessment scores in the training set are used as the output of the demand assessment model, with minimizing the sum of the prediction accuracies of all predicted demand assessment scores as the training goal; training is stopped when the sum of the prediction accuracies converges; the demand assessment model is a naive Bayes model or a support vector machine model.

7. The method for structurally modeling industry knowledge based on a semantically associated knowledge graph according to claim 6 is characterized in that: The training method of the demand construction model includes: A demand construction data set is collected in advance, wherein the demand construction data set includes K groups of demand construction data and predicted demand feature data corresponding to the K groups of demand construction data, where K is a positive integer greater than 0, and the demand construction data includes a demand rationality level, a production parameter set, and a production knowledge graph; the demand construction data set is divided into a training set and a test set, the demand construction data in the training set is used as the input of the demand construction model, and the predicted demand feature data in the training set is used as the output of the demand construction model, with minimizing the sum of the prediction accuracies of all predicted predicted demand feature data as the training goal; training is stopped when the sum of the prediction accuracies converges; the demand construction model is a random forest model or a recurrent neural network model.

8. The method for structurally modeling industry knowledge based on a semantically associated knowledge graph according to claim 7 is characterized in that: The training method of the parameter construction model includes: A parameter construction data set is collected in advance, the parameter construction data set includes H groups of parameter construction data, and a production parameter set corresponding to the H groups of parameter construction data, H is a positive integer greater than 0, and the parameter construction data includes actual demand feature data and a production knowledge graph; the parameter construction data set is divided into a training set and a test set, the parameter construction data in the training set is used as the input of the parameter construction model, and the production parameter set in the training set is used as the output of the parameter construction model, with minimizing the sum of the prediction accuracies of all predicted production parameter sets as the training goal; the training is stopped when the sum of the prediction accuracies converges; the parameter construction model is a CaffeNet model.

9. The method for structurally modeling industry knowledge based on a knowledge graph of semantic association according to claim 8, characterized in that: The industry production data includes production process variables, equipment operating status, production quality indicators and raw material characteristics; the production process variables include temperature, pressure, processing speed, cutting depth and coating thickness; the equipment operation data includes equipment load, current, voltage and vibration frequency; the production quality indicators include pass rate, precision, strength and sealing; the raw material characteristics include composition, specification, ductility and density; the personalized demand content includes expected product type, expected cost, expected pass rate, expected precision and expected delivery cycle.

Images