Steel quality diagnosis method, device, equipment and storage medium

Abstract

The invention provides a steel quality diagnosis method and device, equipment and a storage medium, and the method comprises the steps: obtaining technological parameters and quality parameters of a steel production process, obtaining domain knowledge information, building a quality defect knowledge graph based on the domain knowledge information, and injecting the quality defect knowledge graph into a preset data diagnosis model, thereby obtaining an enhanced diagnosis model; performing predictive diagnosis on the process parameters and the quality parameters based on the enhanced diagnosis model to obtain a predictive diagnosis result, performing reasoning diagnosis on the process parameters and the quality parameters based on the quality defect knowledge graph to obtain a reasoning diagnosis result, and if the predictive diagnosis result and the reasoning diagnosis result have conflicts, judging whether the process parameters and the quality parameters are conflicted or not. If yes, processing according to a preset conflict resolution strategy to obtain a steel quality diagnosis result; according to the invention, based on cooperative work of prediction diagnosis and reasoning diagnosis, diagnosis result differences are processed in combination with a conflict resolution strategy, the accuracy and interpretability of steel quality diagnosis are improved, and the adaptability of a diagnosis system under complex working conditions is enhanced.

Description

Technical Field

[0001] This application relates to the field of production management, and in particular to a method, apparatus, equipment and storage medium for diagnosing steel quality. Background Technology

[0002] As a key raw material, the quality and stability of steel directly affect the overall level of major equipment manufacturing and infrastructure construction. However, steel production involves a long process, tightly coupled procedures, and complex and diverse factors influencing quality. Therefore, steel quality diagnosis has become a core technical challenge. Currently, commonly used diagnostic methods fall into two categories: firstly, numerical analysis of massive amounts of parameters collected during production using machine learning models. However, these models lack an understanding of metallurgical mechanisms and process principles, resulting in a significant decrease in generalization ability when facing new operating conditions, and the diagnostic process struggles to trace the actual physical meaning of the production process; secondly, diagnostic systems built on knowledge graph technology, while conforming to engineering logic, cannot quantify their data analysis capabilities and effective data mining, and the maintenance of knowledge graph technology heavily relies on human experience.

[0003] Currently, the two technical approaches mentioned above often lead to conflicting conclusions in practical applications. Data models may output diagnostic results that contradict fundamental manufacturing principles, while knowledge reasoning systems may suffer from incomplete knowledge coverage, affecting the accuracy of the analysis. Such conflicts and inconsistencies between conclusions limit the application value of diagnostic results, hindering efficient fault location and subsequent maintenance after problem diagnosis. Summary of the Invention

[0004] The purpose of this application is to provide a method, apparatus, equipment, and storage medium for diagnosing steel quality, so as to solve the above-mentioned technical problems.

[0005] This application provides a steel quality diagnosis method, which includes: acquiring process parameters and quality parameters of the steel production process, and acquiring domain knowledge information, including textual content and experiential knowledge describing steel production technology; constructing a quality defect knowledge graph based on the domain knowledge information, and injecting the quality defect knowledge graph into a preset data diagnosis model to obtain an enhanced diagnosis model, wherein the quality defect knowledge graph is used to characterize the process and quality defect content and their interrelationships; performing predictive diagnosis on the process parameters and quality parameters based on the enhanced diagnosis model to obtain predictive diagnosis results, and performing inferential diagnosis on the process parameters and quality parameters based on the quality defect knowledge graph to obtain inferential diagnosis results; if there is a conflict between the predictive diagnosis results and the inferential diagnosis results, processing is performed according to a preset conflict resolution strategy to obtain the steel quality diagnosis result.

[0006] In one embodiment of this application, constructing a quality defect knowledge graph based on the domain knowledge information includes: extracting process information, quality defect information, and related relationships from the text content of the domain knowledge information to form initial triples; and establishing causal relationships and rule constraints on the initial triples based on the empirical knowledge of the domain knowledge information to obtain the quality defect knowledge graph.

[0007] In one embodiment of this application, injecting the quality defect knowledge graph into a preset data diagnostic model to obtain an enhanced diagnostic model includes: acquiring an original training dataset, which is used to train the diagnostic model; fusing the rule constraints in the quality defect knowledge graph and the loss function of the preset data diagnostic model to obtain an initial enhanced model; extracting feature vectors of process information, quality defect information, and correlation relationships in the quality defect knowledge graph, and performing feature fusion on the data features of the original training dataset based on the feature vectors to obtain a joint feature space, wherein the feature fusion includes feature concatenation or weighted summation; and training based on the joint feature space and the initial enhanced model to obtain the enhanced diagnostic model.

[0008] In one embodiment of this application, after predicting and diagnosing the process parameters and quality parameters based on the enhanced diagnostic model to obtain a predicted diagnostic result, and then performing inference and diagnosing the process parameters and quality parameters based on the quality defect knowledge graph to obtain an inference diagnostic result, the method further includes: calculating the model confidence of the predicted diagnostic result as a first confidence, calculating the knowledge credibility of the inference diagnostic result as a second confidence, and calculating the absolute difference between the first confidence and the second confidence; if the absolute difference is greater than or equal to a preset conflict threshold, it is determined that there is a conflict between the predicted diagnostic result and the inference diagnostic result; if the absolute difference is less than the preset conflict threshold, there is no conflict between the predicted diagnostic result and the inference diagnostic result.

[0009] In one embodiment of this application, before calculating the model confidence of the predicted diagnostic result as the first confidence and the knowledge credibility of the inferred diagnostic result as the second confidence, the method further includes: determining the diagnostic type information to which the predicted diagnostic result belongs, and retrieving causal relationship information associated with the diagnostic type from the quality defect knowledge graph; performing logical rationality verification on the predicted diagnostic result based on the causal relationship information; and determining that there is a semantic contradiction between the predicted diagnostic result and any associated causal relationship rule in the causal relationship information if the predicted diagnostic result and the inferred diagnostic result are in conflict.

[0010] In one embodiment of this application, processing according to a preset conflict resolution strategy to obtain a steel quality diagnosis result includes: if the process parameters do not meet preset process parameter quality conditions, then the inference diagnosis result is determined as the steel quality diagnosis result. The preset process parameter quality conditions include the integrity of the process parameters being lower than a preset integrity threshold, and the value distribution of the process parameters in historical data exceeding the coverage of historical training data. The integrity of the process parameters refers to the ratio of the actual number of process parameters collected to the number that should be collected; or, a weighted fusion is performed based on the first confidence level of the predicted diagnosis result and the second confidence level of the inference diagnosis result, and the predicted diagnosis result or inference diagnosis result with the highest comprehensive confidence level is taken as the steel quality diagnosis result.

[0011] In one embodiment of this application, after obtaining the steel quality diagnosis result, the method further includes: if there is a conflict between the predicted diagnosis result and the inferred diagnosis result, the selection information of the conflict resolution strategy type, the weight configuration parameters of the steel quality diagnosis result determination process, and the steel quality diagnosis result are recorded as conflict cases, so as to update the quality defect knowledge graph based on the conflict cases.

[0012] This application embodiment also provides a steel quality diagnostic device, which includes: a diagnostic parameter acquisition module, used to acquire process parameters and quality parameters of the steel production process, and acquire domain knowledge information, the domain knowledge information including text content and experiential knowledge describing steel production technology; a diagnostic model training module, used to construct a quality defect knowledge graph based on the domain knowledge information, and inject the quality defect knowledge graph into a preset data diagnostic model to obtain an enhanced diagnostic model, the quality defect knowledge graph being used to characterize the process and quality defect content and their interrelationships; a diagnosis and arbitration module, used to perform predictive diagnosis on the process parameters and quality parameters based on the enhanced diagnostic model to obtain a predictive diagnosis result, and to perform inference diagnosis on the process parameters and quality parameters based on the quality defect knowledge graph to obtain an inference diagnosis result; if there is a conflict between the predictive diagnosis result and the inference diagnosis result, the conflict is resolved according to a preset conflict resolution strategy to obtain a steel quality diagnostic result.

[0013] This application also provides an electronic device, including: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the electronic device to implement the steel quality diagnosis method as described in any of the above embodiments.

[0014] This application also provides a computer-readable storage medium storing computer-readable instructions thereon, which, when executed by a computer's processor, cause the computer to perform the steel quality diagnosis method as described in any of the above embodiments.

[0015] The beneficial effects of this application are as follows: This application provides a steel quality diagnosis method, apparatus, equipment, and storage medium. It acquires process and quality parameters from the steel production process, obtains domain knowledge information, constructs a quality defect knowledge graph based on the domain knowledge information, and injects it into a preset data diagnosis model to obtain an enhanced diagnosis model. Based on the enhanced diagnosis model, it performs predictive diagnosis on the process and quality parameters to obtain predictive diagnosis results, and performs inference diagnosis on the process and quality parameters based on the quality defect knowledge graph to obtain inference diagnosis results. If there is a conflict between the predictive and inference diagnosis results, it is handled according to a preset conflict resolution strategy to obtain the steel quality diagnosis result. This application achieves collaborative work between predictive and inference diagnosis by integrating domain knowledge information to construct a quality defect knowledge graph and injecting it into the diagnosis model. Combined with conflict resolution strategies to handle differences in diagnosis results, it improves the accuracy and interpretability of steel quality diagnosis and enhances the adaptability of the diagnosis system under complex working conditions.

[0016] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. In the drawings: Figure 1 This is a schematic diagram illustrating an exemplary system architecture as shown in an exemplary embodiment of this application; Figure 2 An exemplary embodiment of this application illustrates a flowchart of a steel quality diagnosis method; Figure 3 This is a schematic diagram illustrating a specific steel quality diagnosis method implemented in an exemplary embodiment of this application; Figure 4 This is a schematic diagram of a steel quality diagnostic device shown in an exemplary embodiment of this application; Figure 5 This is a schematic diagram of the structure of a computer system for an electronic device, as illustrated in an exemplary embodiment of this application. Detailed Implementation

[0018] The embodiments of this application will be described below with reference to the accompanying drawings and specific examples. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be understood that the preferred embodiments are only for illustrating this application and are not intended to limit the scope of protection of this application.

[0019] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the illustrations only show the components related to this application and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0020] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the present application. However, it will be apparent to those skilled in the art that embodiments of the present application may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the present application.

[0021] The term "and / or" used in this application describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The character " / " generally indicates that the related objects before and after it are in an "or" relationship.

[0022] In implementing related technologies, data-driven models perform numerical analysis based on production parameters. However, due to a lack of internalization of metallurgical mechanisms, their output often contradicts basic manufacturing principles under new operating conditions. Knowledge reasoning systems, built on knowledge graphs, conform to engineering logic, but are limited by incomplete knowledge coverage and insufficient data analysis capabilities. This leads to conflicts between the two types of diagnostic conclusions, reducing the reliability of diagnostic results, hindering fault location, and affecting the accuracy of subsequent maintenance decisions. If this conflict is not resolved, the lack of traceability in the diagnostic process will leave maintenance measures without a basis, and product quality defects may cause a significant decrease in the precision of equipment manufacturing and safety hazards in infrastructure construction, continuously damaging the stability of the overall production system.

[0023] Based on the aforementioned technical problems and characteristics, this application proposes the following technical solution: By acquiring the process parameters and quality parameters of the steel production process and obtaining domain knowledge information, a quality defect knowledge graph is constructed based on the domain knowledge information and injected into a preset data diagnostic model to obtain an enhanced diagnostic model. Based on the enhanced diagnostic model, the process parameters and quality parameters are predicted and diagnosed to obtain predicted diagnostic results. Furthermore, based on the quality defect knowledge graph, the process parameters and quality parameters are inferred and diagnosed to obtain inferred diagnostic results. If there is a conflict between the predicted diagnostic results and the inferred diagnostic results, it is processed according to a preset conflict resolution strategy to obtain the steel quality diagnostic results. Through the collaborative operation of the dual diagnostic paths, it is ensured that the predicted diagnostic results have data quantification and analysis capabilities, while the inferred diagnostic results conform to the principles of metallurgical processes, effectively avoiding the conclusion bias caused by a single diagnostic path.

[0024] Please see Figure 1 , Figure 1 This is a schematic diagram illustrating an exemplary system architecture as shown in an exemplary embodiment of this application.

[0025] Reference Figure 1 As shown, the system architecture may include a steel production database 110 and a computer device 120. The computer device 120 obtains process and quality parameters of the steel production process from the steel production database 110, and acquires domain knowledge information. Based on the domain knowledge information, it constructs a quality defect knowledge graph and injects it into a preset data diagnostic model to obtain an enhanced diagnostic model. The quality defect knowledge graph is used to characterize the process and quality defect content and their interrelationships. Based on the enhanced diagnostic model, the process and quality parameters are predicted and diagnosed to obtain predicted diagnostic results. The process and quality parameters are also inferred and diagnosed based on the quality defect knowledge graph to obtain inferred diagnostic results. If there is a conflict between the predicted and inferred diagnostic results, it is processed according to a preset conflict resolution strategy to obtain the steel quality diagnostic result. The steel production database 110 is used to collect and temporarily store the process and quality parameters of the steel production process and transmit them to the computer device 120. The computer device 120 refers to a computing power support terminal device used to support the program implementation environment for the steel quality diagnostic method, including but not limited to microcomputers, tablets, industrial control computers, server clusters, and cloud servers.

[0026] This technical solution can be further discussed and analyzed in multiple steps; for specific steps, please refer to [link / reference needed]. Figure 2 , Figure 2 This is a flowchart illustrating an exemplary embodiment of a steel quality diagnosis method. This steel quality diagnosis method can be executed in implementation environments supported by various operating systems, and no specific limitation is made to the implementation environment herein. (Refer to...) Figure 2As shown, the flowchart of this steel quality diagnosis method includes at least steps S210 to S240, which are described in detail below: In step S210, process parameters and quality parameters of the steel production process are obtained, and domain knowledge information is acquired.

[0027] In one embodiment of this application, the aforementioned domain knowledge information includes textual content and experiential knowledge used to describe steel production technology. Specifically, it can be achieved by collecting industry technical specification documents, organizing production operation logs, or recording key process points through expert interviews. For example, process parameter definitions can be extracted from metallurgical handbooks, and empirical rules can be summarized based on historical fault reports to provide comprehensive process principle support and avoid the diagnostic process from deviating from actual production logic.

[0028] In one embodiment of this application, the process parameters and quality parameters of the steel production process are obtained based on the steel plant's steel production database, covering production processes such as steelmaking, continuous casting, heating furnace, and rolling line. The process parameters can be further divided into steel grade specifications, chemical composition, heating furnace parameters, and rolling process parameters. The quality parameters further include quality labels. The chemical composition includes constituent elements and their contents such as Al, Als, As, B, C, Ca, Cr, Cu, Mn, Mo, N, Nb, Ni, P, S, Si, Ti, and V. The heating furnace parameters include parameters such as the temperature and duration of each section, such as the furnace inlet temperature, preheating section temperature, heating section temperature, soaking section temperature, furnace outlet temperature, heating duration, soaking duration, and furnace dwell time. The rolling process includes, but is not limited to, information such as current, elongation, speed, and temperature.

[0029] In step S220, a quality defect knowledge graph is constructed based on domain knowledge information, and the quality defect knowledge graph is injected into a preset data diagnostic model to obtain an enhanced diagnostic model.

[0030] In one embodiment of this application, the aforementioned quality defect knowledge graph is used to characterize the process and the content of quality defects and their interrelationships. Specifically, it refers to a knowledge base that expresses the relationship between the process and quality defects in a structured form, so as to establish a traceable process logic basis to support the reasoning and diagnosis process.

[0031] In one embodiment of this application, constructing a quality defect knowledge graph based on domain knowledge information includes extracting process information, quality defect information, and related relationships to form initial triples. Then, based on experiential knowledge from the domain knowledge information, causal relationships and rule constraints are established on the initial triples to obtain the quality defect knowledge graph. Specifically, unstructured text data is parsed using text mining techniques to form initial triples. Named entity recognition and relation extraction techniques from natural language processing are used for triple extraction and construction, which can be based on tools such as Graphusion, KGGen, and DeepKE. Establishing causal relationships and rule constraints on the initial triples involves introducing the practical experience of domain experts to logically enhance the data. Verification and supplementation are performed using expert systems or rule engines to supplement metallurgical mechanisms and process rules not explicitly stated in the text, thereby improving the reasoning ability of the knowledge graph. Specifically, process nodes, quality nodes, and defect nodes are used as entities, and process-quality association, quality-defect causality, and process-defect propagation are used as relation edges to construct the triple structure.

[0032] In the embodiments of this application, process information, quality defect information and related relationships are first extracted from the text content of domain knowledge information to form initial triples, and a high-coverage basic data layer is constructed. Subsequently, based on the empirical knowledge of domain knowledge information, causal relationships and rule constraints are established on the initial triples to form a structured knowledge system with causal reasoning ability. The knowledge graph constructed based on this is both comprehensive and logically rigorous, so that the logical rationality can be effectively verified in the diagnostic process, and reasoning bias caused by knowledge gaps can be avoided.

[0033] In one embodiment of this application, injecting a quality defect knowledge graph into a preset data diagnostic model to obtain an enhanced diagnostic model includes obtaining an original training dataset, which is used to train the diagnostic model, and fusing the rule constraints in the quality defect knowledge graph with the loss function of the preset data diagnostic model to obtain an initial enhanced model.

[0034] The original training dataset is the basic data set used to train the diagnostic model, which can be historical data of steel production or real-time collected process parameter data; the fusion of rule constraints and loss function integrates the logical rules in the knowledge graph into the model optimization objective, which can be achieved by adding regularization terms.

[0035] In one embodiment of this application, feature vectors of process information, quality defect information, and related relationships are extracted from the quality defect knowledge graph. Based on these feature vectors, feature fusion is performed on the data features of the original training dataset to obtain a joint feature space. Feature fusion includes feature concatenation or weighted summation. An enhanced diagnostic model is then trained based on the joint feature space and an initial enhancement model. Specifically, feature vector extraction and feature fusion can employ graph embedding technology to obtain structured information from the knowledge graph and combine it with data features to enhance the expressive power of the feature space. The joint feature space training specifically employs end-to-end training to simultaneously strengthen data-driven prediction capabilities and the consistency of knowledge reasoning logic.

[0036] In the embodiments of this application, the diagnostic model's ability to integrate domain knowledge is enhanced, enabling the model to simultaneously absorb data-driven patterns and domain knowledge rules during the training phase. This effectively reduces the conflict between predicted diagnostic results and inferred diagnostic results, and improves the reliability and interpretability of steel quality diagnostic results.

[0037] In one embodiment of this application, a dynamic enhancement mechanism for the knowledge graph is also included. This mechanism enhances the uncertainty assessment of production samples using an enhanced diagnostic model, identifies samples requiring confirmation whose diagnostic confidence is below a set threshold, and pushes these samples and their corresponding predicted diagnostic results for review. After review and confirmation, the aforementioned quality defect knowledge graph is supplemented and corrected based on the review results. The uncertainty assessment includes, but is not limited to, calculating the variance of multiple predictions made by the enhanced diagnostic model for the same input sample, evaluating the probability distribution entropy of predicted diagnostic results across different diagnostic types, and analyzing the feature space distance between the current input sample and the training dataset. The supplementation and correction of the aforementioned quality defect knowledge graph includes, but is not limited to, adding newly confirmed knowledge to the quality defect knowledge graph in the form of triples, correcting existing rules or relationships in the knowledge graph that have been proven inaccurate by review, updating the confidence weights of entities and relationships in the knowledge graph, and establishing a traceability association between newly added knowledge and its source samples.

[0038] In step S230, the process parameters and quality parameters are predicted and diagnosed based on the enhanced diagnostic model to obtain the predicted diagnostic results, and the process parameters and quality parameters are inferred and diagnosed based on the quality defect knowledge graph to obtain the inferred diagnostic results.

[0039] In one embodiment of this application, process parameters are processed based on a temporal neural network to extract process features, and the process features and quality parameters are jointly input into an enhanced diagnostic model to output quality classification results and defect probability predictions, thereby obtaining a predicted diagnostic result. The aforementioned temporal neural network is a long short-term memory network with an attention mechanism, used for feature extraction and key process identification of multi-source process parameters.

[0040] In one embodiment of this application, the current process parameters and quality parameters are mapped to query paths in a knowledge graph. Multi-hop reasoning is performed on the query paths based on a graph neural network to trace the root causes of potential defects and generate explanatory diagnostic results, which are then used as inference diagnostic results.

[0041] In one embodiment of this application, after obtaining the predicted diagnostic result and the inferred diagnostic result, the model confidence of the predicted diagnostic result is calculated as the first confidence, and the knowledge credibility of the inferred diagnostic result is calculated as the second confidence, and the absolute difference between the first confidence and the second confidence is calculated. The model confidence is a quantitative indicator of the data-driven diagnostic result based on the probability distribution entropy value of the model output, the statistical confidence interval of the predicted result, or the cross-validation consistency index based on historical data, used to objectively reflect the degree of certainty of process parameters at the statistical level. The knowledge credibility is a reliability assessment indicator of the knowledge inferred diagnostic result, calculated based on the weighted comprehensive value of the rule matching completeness score, the logical self-consistency of the causal chain, and the coverage of expert experience. The absolute difference is a numerical difference measure between the first confidence and the second confidence.

[0042] If the absolute difference is greater than or equal to the preset conflict threshold, then the predicted diagnosis result and the inferred diagnosis result are determined to be in conflict.

[0043] If the absolute difference is less than the preset conflict threshold, there is no conflict between the predicted diagnosis result and the inferred diagnosis result.

[0044] The aforementioned preset conflict threshold can be calculated and determined based on the statistical distribution analysis of historical conflict cases, the dynamic adjustment mechanism set by domain experts, and the fixed threshold determined by the optimization algorithm.

[0045] In the embodiments of this application, by converting the model confidence of the predicted diagnostic results and the knowledge credibility of the inferred diagnostic results into quantitative numerical indicators, a comparison mechanism is established between the absolute difference between the two and a preset conflict threshold, forming a closed-loop conflict determination logic. First, the statistical reliability of the data-driven diagnosis is calculated based on the model confidence, while the logical completeness of the knowledge reasoning is assessed based on the knowledge credibility. Then, the absolute difference directly reflects the degree of deviation between the two types of diagnostic results. When the absolute difference exceeds the preset conflict threshold, a significant conflict is determined, and subsequent processing is triggered; when the absolute difference is within the threshold range, the consistency of the diagnostic results is confirmed, and a conclusion is directly output. This approach avoids the limitations of relying solely on a single diagnostic method, ensuring that conflict determination is initiated only when the discrepancy exceeds an acceptable range, effectively preventing excessive intervention in minor differences, and reducing unnecessary conflict handling processes.

[0046] In one embodiment of this application, before calculating the first confidence level and the second confidence level, the diagnosis type information to which the predicted diagnosis result belongs is determined, and the causal relationship information associated with the diagnosis type is retrieved from the quality defect knowledge graph. Based on the causal relationship information, the logical rationality of the predicted diagnosis result is verified.

[0047] Specifically, determining the diagnostic type information of the predicted diagnostic result can be achieved by using a keyword matching-based classification method or a pre-trained classification model to identify the specific category of the predicted diagnostic result; retrieving causal relationship information related to the diagnostic type from the quality defect knowledge graph includes using graph database queries or semantic similarity matching to retrieve relevant process rules based on the diagnostic type; the above logical rationality verification includes comparing the predicted result with domain knowledge rules based on logical reasoning algorithms to verify whether the predicted result conforms to the principles of metallurgical processes.

[0048] If the predicted diagnostic result has a semantic contradiction with any causal relationship rule associated with the causal relationship information, then the predicted diagnostic result and the inferred diagnostic result are determined to be in conflict. The existence of such semantic contradictions can be determined based on a semantic contradiction detection algorithm, avoiding serious logical errors that may be overlooked due to small numerical differences.

[0049] In the embodiments of this application, the logical rationality of the predicted diagnostic results is verified before the confidence level is calculated, effectively identifying diagnostic results that violate the principles of metallurgical processes. Even if the confidence level difference is small, the conflict can be accurately determined, thereby improving the reliability of steel quality diagnosis.

[0050] In step S240, if there is a conflict between the predicted diagnosis result and the inferred diagnosis result, the conflict is resolved according to the preset conflict resolution strategy to obtain the steel quality diagnosis result.

[0051] In one embodiment of this application, if the process parameters do not meet the preset process parameter quality conditions, the inference diagnosis result is used to determine the steel quality diagnosis result.

[0052] The aforementioned preset process parameter quality conditions include process parameter completeness falling below a preset completeness threshold, and the distribution of process parameter values ​​in historical data exceeding the coverage range of historical training data. Process parameter completeness refers to the ratio of the actual number of process parameters collected to the number that should have been collected. These preset process parameter quality conditions serve as a benchmark for evaluating process parameter data quality and can be determined using indicators such as process parameter completeness and value distribution range. Process parameter completeness is a quantitative indicator of the completeness of process parameter collection, determined by calculating the ratio of the actual number of collected parameters to the number that should have been collected. Value distribution exceeding the coverage range of historical training data means that the process parameter values ​​under the current operating conditions exceed the historical data distribution range during model training; outliers can be detected using statistical methods.

[0053] Alternatively, a weighted fusion can be performed based on the first confidence level of the predicted diagnostic result and the second confidence level of the inferred diagnostic result, with the predicted or inferred diagnostic result having the highest overall confidence level as the steel quality diagnostic result. Specifically, the confidence-based weighted fusion uses a linear weighting algorithm to allocate weights according to the first confidence level of the predicted diagnostic result and the second confidence level of the inferred diagnostic result, taking the result with the highest overall confidence level as the final diagnostic output. This fully utilizes the confidence index to objectively assess the reliability of each method, avoiding the limitations of relying solely on models or knowledge, and dynamically balancing the advantages of data-driven and knowledge-driven methods.

[0054] In one embodiment of this application, after obtaining the steel quality diagnosis result, if there is a conflict between the predicted diagnosis result and the inferred diagnosis result, the selection information of the conflict resolution strategy type, the weight configuration parameters of the steel quality diagnosis result determination process, and the steel quality diagnosis result are recorded as conflict cases, so as to update the quality defect knowledge graph based on the conflict cases. The conflict cases are structured datasets that integrate conflict-related information, providing complete multi-dimensional historical data for knowledge graph updates.

[0055] In one embodiment of this application, when a conflict is confirmed, the selection information of the conflict resolution strategy type, weight configuration parameters, and steel quality diagnosis results are automatically collected and integrated into a structured conflict case, which is used in the updating process of the quality defect knowledge graph. By analyzing the causal contradictions or rule coverage gaps in the case, the rule constraints of the knowledge graph are modified in a targeted manner or new associations are added, thereby realizing the dynamic optimization and adaptive capability improvement of the knowledge graph.

[0056] Please refer to Figure 3 , Figure 3This is an exemplary embodiment of the present application illustrating a specific steel quality diagnosis method. In one specific embodiment, after the diagnosis process begins, production data is first collected, and process experience is gathered to form a knowledge graph, consistent with the quality defect knowledge graph in the above embodiment. A knowledge-injected data model is first constructed and trained. Knowledge is incorporated into the features and loss function of the initial enhancement model to constrain it. Firstly, rule constraints from the knowledge graph, such as logical rules or domain prior knowledge, are incorporated into the loss function. Specifically, this includes feature concatenation and weighted summation. The feature concatenation method concatenates the entity / relation embedding vector with the original feature vector column-wise to form a new feature vector. The weighted summation involves weighting the entity / relation embedding vector and the original feature vector. By assigning different weights to different features, their relative importance in the fusion can be controlled. Then, the entity / relation embedding from the knowledge graph is fused with the original features of the data model to enhance the model's perception of semantic consistency. Finally, the uncertainty sampling of the data model is used to identify samples that may trigger conflicts and actively feed them back to the diagnosis system for manual review. For samples that may trigger conflicts, information such as entities, relationships, and attributes is extracted and integrated with the current knowledge graph database to ensure the consistency of knowledge. This information is then reviewed by domain experts to ensure the accuracy and reliability of the knowledge. The confidence level of the diagnostic results is then used to determine whether a conflict may be triggered. If the confidence level based on the data model is significantly higher than that based on the knowledge graph, it is considered that a conflict may be triggered and is then fed back to the diagnostic system for manual review.

[0057] In one specific embodiment of this application, conflict detection involves consistency verification between quality diagnostic results based on a knowledge graph and quality diagnostic results based on a data model. First, the predicted diagnostic results are verified based on the quality defect knowledge graph. For example, if the data model attributes the generation of iron oxide scale to the heating time of the furnace, and this heating time is relatively short in historical data (i.e., the data model believes that "the heating time of the furnace is too short, leading to the generation of iron oxide scale"), but the quality defect knowledge graph records that "the heating time of the furnace is too long, which may lead to the generation of iron oxide scale," then a conflict is triggered.

[0058] In one specific embodiment of this application, a confidence level is added to the model that predicts the diagnostic results, and a confidence weight is added to the quality defect knowledge graph rules. The difference between the two confidence levels is compared, and the input data, model version, and knowledge graph of the conflicting samples are recorded.

[0059] In one specific embodiment of this application, for conflicts detected or judged manually, dynamic decision-making and feedback loops can be implemented using various methods to obtain a unique diagnostic conclusion on the system. Firstly, a priority strategy can be used, including setting static priorities based on the characteristics of the diagnostic domain; scenarios with high data uncertainty or reliance on expert experience generally prioritize knowledge graphs. Secondly, dynamic weight fusion can be used, where weighted results are calculated based on confidence levels, which can be characterized as follows: Final Score = α Model_Confidence+β KG_Confidence Wherein, Final Score is the final confidence score, α is the weight of the first confidence score, β is the weight of the second confidence score, Model_Confidence is the model confidence score for predicting the diagnostic results, i.e., the first confidence score, and KG_Confidence is the knowledge confidence score for inferring the diagnostic results, i.e., the second confidence score.

[0060] In one specific embodiment of this application, the final step is conflict resolution based on expert experience. When a conflict occurs, process experts are introduced to verify and determine the diagnostic results.

[0061] After conflict resolution, the diagnostic results are output. The resolved diagnostic results are then used to update the knowledge graph or trigger the data model retraining. Conflict cases are labeled and stored in the knowledge base to optimize subsequent conflict detection rules.

[0062] This application provides a steel quality diagnosis method, apparatus, equipment, and storage medium. It acquires process and quality parameters from the steel production process, along with domain knowledge information. Based on this domain knowledge, a quality defect knowledge graph is constructed and injected into a pre-defined data diagnosis model to obtain an enhanced diagnosis model. The enhanced diagnosis model is then used to predict and diagnose the process and quality parameters, yielding predicted diagnosis results. Furthermore, based on the quality defect knowledge graph, the process and quality parameters are used for inference and diagnosis, resulting in inference diagnosis results. If there is a conflict between the predicted and inference diagnosis results, a pre-defined conflict resolution strategy is employed to resolve the conflict, ultimately yielding the final steel quality diagnosis result. This application achieves collaborative work between predictive and inference diagnosis by integrating domain knowledge information to construct a quality defect knowledge graph and injecting it into the diagnosis model. The conflict resolution strategy addresses discrepancies in the diagnosis results, improving the accuracy and interpretability of steel quality diagnosis and enhancing the system's adaptability under complex operating conditions.

[0063] The following describes an embodiment of the apparatus described in this application, which can be used to execute the steel quality diagnosis method described in the above embodiments of this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the steel quality diagnosis method described above in this application.

[0064] Figure 4 This is a schematic diagram illustrating a steel quality diagnostic device according to an exemplary embodiment of this application. The device can be applied to… Figure 1 The method described is implemented in a device that has the necessary conditions for execution. This embodiment does not impose specific limitations on the devices to which the device is applicable.

[0065] like Figure 4 As shown, the exemplary steel quality diagnostic device includes: a diagnostic parameter acquisition module 401, a diagnostic model training module 402, and a diagnostic and arbitration module 403.

[0066] The diagnostic parameter acquisition module 401 is used to acquire process parameters and quality parameters of the steel production process, and to acquire domain knowledge information, including textual content and experiential knowledge describing steel production technology. The diagnostic model training module 402 is used to construct a quality defect knowledge graph based on the domain knowledge information, and to inject the quality defect knowledge graph into a preset data diagnostic model to obtain an enhanced diagnostic model. The quality defect knowledge graph is used to characterize the process and quality defect content and their interrelationships. The diagnosis and arbitration module 403 is used to predict and diagnose the process parameters and quality parameters based on the enhanced diagnostic model to obtain the predicted diagnostic results, and to perform inference and diagnosis on the process parameters and quality parameters based on the quality defect knowledge graph to obtain the inference diagnostic results. If there is a conflict between the predicted diagnostic results and the inference diagnostic results, it is processed according to a preset conflict resolution strategy to obtain the steel quality diagnostic results.

[0067] Embodiments of this application also provide an electronic device, including: one or more processors; and a storage device for storing one or more programs, which, when executed by one or more processors, cause the electronic device to implement the steel quality diagnosis method provided in the above embodiments.

[0068] Figure 5 This is a schematic diagram illustrating the structure of a computer system for an electronic device, as shown in an exemplary embodiment of this application. It should be noted that... Figure 5 The computer system 500 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0069] like Figure 5As shown, the computer system 500 includes a Central Processing Unit (CPU) 501, which can perform various appropriate actions and processes, such as executing the methods described in the above embodiments, based on programs stored in Read-Only Memory (ROM) 502 or programs loaded from storage into Random Access Memory (RAM) 503. The RAM 503 also stores various programs and data required for system operation. The CPU 501, ROM 502, and RAM 503 are interconnected via a bus. An I / O interface 505 is also connected to the bus 504, where the I / O interface 505 refers to an input / output interface.

[0070] The following components are connected to I / O interface 505: input section 506 including keyboard, mouse, etc.; output section 507 including cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; storage section 508 including hard disk, etc.; and communication section 509 including network interface card, such as LAN (Local Area Network) card, modem, etc. The communication section performs communication processing via a network such as the Internet. A drive is also connected to I / O interface 505 as needed. Removable media 511, such as disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 510 as needed so that computer programs read from them can be installed into storage section 508 as needed.

[0071] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by central processing unit (CPU) 501, it performs various functions defined in the system of this application.

[0072] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0073] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0074] In the corresponding figures of the above embodiments, connecting lines can represent the connection relationship between various components, indicating more constitutive signal paths and / or one or more ends of some lines having arrows to indicate the main information flow direction. Connecting lines are an identifier and are not a limitation on the scheme itself, but rather, using these lines in conjunction with one or more exemplary embodiments helps to more easily connect circuits or logic units. Any signal represented (determined by design requirements or preferences) can actually include one or more signals that can be transmitted in any direction and can be implemented in any suitable type of signal scheme.

[0075] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0076] Another aspect of this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method as described above. This computer-readable storage medium may be included in the electronic device described in the above embodiments, or it may exist independently and not assembled into the electronic device.

[0077] Embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the steel quality diagnosis method as described in any of the above embodiments.

[0078] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0079] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the method according to the embodiments of this application.

[0080] This application can be used in a wide range of general-purpose or special-purpose computing system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0081] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0082] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A method for diagnosing steel quality, characterized in that, The steel quality diagnosis method includes: The process parameters and quality parameters of the steel production process are obtained, and domain knowledge information is also obtained, including textual content and empirical knowledge used to describe steel production technology. A quality defect knowledge graph is constructed based on the domain knowledge information, and the quality defect knowledge graph is injected into a preset data diagnostic model to obtain an enhanced diagnostic model. The quality defect knowledge graph is used to characterize the process and quality defect content and their interrelationships. Based on the enhanced diagnostic model, the process parameters and quality parameters are predicted and diagnosed to obtain the predicted diagnostic results. Based on the quality defect knowledge graph, the process parameters and quality parameters are inferred and diagnosed to obtain the inferred diagnostic results. If there is a conflict between the predicted diagnostic results and the inferred diagnostic results, the conflict will be resolved according to the preset conflict resolution strategy to obtain the steel quality diagnostic results.

2. The steel quality diagnosis method according to claim 1, characterized in that, Constructing a quality defect knowledge graph based on the aforementioned domain knowledge information includes: Extract process information, quality defect information, and correlations from the text content of the domain knowledge information to form an initial triplet; Based on the empirical knowledge of the domain information, causal relationships and rule constraints are established for the initial triples to obtain a quality defect knowledge graph.

3. The steel quality diagnosis method according to claim 2, characterized in that, Injecting the quality defect knowledge graph into a preset data diagnostic model yields an enhanced diagnostic model, including: Obtain the original training dataset, which is used to train the diagnostic model; The rule constraints in the quality defect knowledge graph and the loss function of the preset data diagnosis model are fused to obtain the initial enhancement model; Extract feature vectors of process information, quality defect information and related relationships from the quality defect knowledge graph, and perform feature fusion on the data features of the original training dataset based on the feature vectors to obtain a joint feature space. The feature fusion includes feature concatenation or weighted summation. An enhanced diagnostic model is obtained by training based on the joint feature space and the initial enhancement model.

4. The steel quality diagnosis method according to claim 1, characterized in that, Based on the enhanced diagnostic model, predictive diagnostics are performed on the process parameters and quality parameters to obtain predictive diagnostic results. Then, based on the quality defect knowledge graph, inference diagnostics are performed on the process parameters and quality parameters to obtain inference diagnostic results. The process parameters and quality parameters are then further analyzed, including: The model confidence score of the predicted diagnostic result is calculated as the first confidence score, the knowledge confidence score of the inferred diagnostic result is calculated as the second confidence score, and the absolute difference between the first confidence score and the second confidence score is calculated. If the absolute difference is greater than or equal to a preset conflict threshold, then it is determined that there is a conflict between the predicted diagnosis result and the inferred diagnosis result. If the absolute difference is less than a preset conflict threshold, there is no conflict between the predicted diagnosis result and the inferred diagnosis result.

5. The steel quality diagnosis method according to claim 4, characterized in that, Before calculating the model confidence score of the predicted diagnostic result as the first confidence score and the knowledge confidence score of the inferred diagnostic result as the second confidence score, the method further includes: Determine the diagnosis type information to which the predicted diagnosis result belongs, and retrieve the causal relationship information associated with the diagnosis type from the quality defect knowledge graph; Based on the causal relationship information, the logical rationality of the predicted diagnostic results is verified; If the predicted diagnostic result has a semantic contradiction with any of the associated causal relationship rules in the causal relationship information, then it is determined that the predicted diagnostic result and the inferred diagnostic result are in conflict.

6. The steel quality diagnosis method according to any one of claims 1-5, characterized in that, Based on the preset conflict resolution strategy, the steel quality diagnosis results include: If the process parameters do not meet the preset process parameter quality conditions, the inference diagnosis result will be used to determine the steel quality diagnosis result. The preset process parameter quality conditions include the integrity of the process parameters being lower than the preset integrity threshold, and the value distribution of the process parameters in historical data exceeding the coverage of historical training data. The integrity of the process parameters refers to the ratio of the actual number of process parameters collected to the number that should be collected. Alternatively, a weighted fusion can be performed based on the first confidence level of the predicted diagnostic result and the second confidence level of the inferred diagnostic result, and the predicted diagnostic result or the inferred diagnostic result with the highest comprehensive confidence level can be used as the steel quality diagnostic result.

7. The steel quality diagnosis method according to any one of claims 1-5, characterized in that, After obtaining the steel quality diagnosis results, the following is also included: If there is a conflict between the predicted diagnosis result and the inferred diagnosis result, the selection information of the conflict resolution strategy type, the weight configuration parameters of the steel quality diagnosis result determination process, and the steel quality diagnosis result are recorded as conflict cases, so as to update the quality defect knowledge graph based on the conflict cases.

8. A steel quality diagnostic device, characterized in that, The steel quality diagnostic device includes: The diagnostic parameter acquisition module is used to acquire process parameters and quality parameters of the steel production process, and to acquire domain knowledge information, which includes text content and empirical knowledge describing steel production technology. The diagnostic model training module is used to construct a quality defect knowledge graph based on the domain knowledge information, and inject the quality defect knowledge graph into a preset data diagnostic model to obtain an enhanced diagnostic model. The quality defect knowledge graph is used to characterize the process and quality defect content and their interrelationships. The diagnosis and arbitration module is used to perform predictive diagnosis on the process parameters and quality parameters based on the enhanced diagnosis model to obtain the predicted diagnosis results, and to perform inference diagnosis on the process parameters and quality parameters based on the quality defect knowledge graph to obtain the inference diagnosis results; if there is a conflict between the predicted diagnosis results and the inference diagnosis results, the conflict is resolved according to a preset conflict resolution strategy to obtain the steel quality diagnosis results.

9. An electronic device, characterized in that, It includes a processor, a memory, and a communication bus; the communication bus is used to connect the processor and the memory; the processor is used to execute a computer program stored in the memory to implement the steel quality diagnosis method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that enables the computer to perform the steel quality diagnosis method as described in any one of claims 1-7.

Images