Modeling and analysis method and system for correlation between glaze wear resistance and process parameters

Abstract

The present invention discloses a modeling and analysis method and system for the correlation between the wear resistance of glaze and process parameters, and relates to the field of data processing technology. It includes model establishment: establishing an analysis model to reflect the correlation between the wear resistance of glaze and process parameters. The present invention sets a multi-level analysis model, which can facilitate users to obtain corresponding parameter data in the analysis model according to their own wear resistance requirements, so that users can understand the correlation between wear resistance and process parameters, and produce products of corresponding requirements according to parameter data, which is conducive to improving work efficiency. Through the set update method, a single analysis model can be optimized and updated. Through the unique multi-level model design and the update method, a single model can be optimized and updated, which can reduce the amount of data processing when training, updating and using the model to improve the response speed and ensure the efficiency of product production. The parameter data does not represent fixed control parameters, which is conducive to the application of different process conditions.

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to a correlation modeling and analysis method and system for glaze wear resistance and process parameters. Background Art

[0002] Glaze is an important coating that enhances the surface properties of materials such as ceramics and porcelain, and its wear resistance depends on the combined effect of multiple process parameters. During the production and application process, understanding the relationship between these parameters is crucial to optimizing glaze performance and improving product quality.

[0003] Patent publication number CN117217134A discloses a method, device, and storage medium for preparing a wear-resistant circuit board, the method comprising the steps of: obtaining daily use status parameters of the circuit board; calculating expected wear resistance parameters based on the daily use status parameters; obtaining existing specification parameters of the circuit board; and preparing the circuit board based on the expected wear resistance parameters and the existing specification parameters. The method, device, and storage medium for preparing a wear-resistant circuit board provided in the embodiment of the present application can calculate the expected wear resistance parameters with reference to the daily use status parameters of the circuit board, and improve the material type and structure of the circuit board with reference to the difference between the existing specification parameters of the circuit board and the expected wear resistance parameters, thereby ensuring that the circuit board is more wear-resistant and applicable to more complex environments.

[0004] When preparing articles, in order to adjust the performance parameters of the articles to the target parameters, it is often necessary to understand the relationship between the performance of the articles and the production parameters. By understanding the relationship between the performance of the articles and the production parameters, in the subsequent production process, the performance of the articles can be adjusted by adjusting the production parameters, so as to produce articles that meet the actual use requirements. In the above-mentioned and similar methods, the correlation between the wear resistance and the parameters is not clear enough, and most models are more complicated to update their correlations. A large amount of data processing is required during the update, which is inefficient. The failure to update the optimized correlation will affect the subsequent preparation quality and efficiency of the articles. In the prior art, under different process conditions, the model may need to be retrained, that is, the model cannot be well generalized to the new production environment, which is not convenient for saving production costs and improving production quality and efficiency. Therefore, the present invention is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for modeling and analyzing the correlation between glaze wear resistance and process parameters, so as to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a correlation modeling and analysis method between glaze wear resistance and process parameters, the method comprising:

[0007] Model establishment: Establish an analytical model to reflect the correlation between glaze wear resistance and process parameters;

[0008] Data acquisition: Obtain production data and experimental data, integrate production data and experimental data to obtain basic data, which includes wear resistance and process parameters corresponding to wear resistance:

[0009] Linear judgment: Based on the basic data, the relationship between the wear resistance and the change of process parameters is judged by the linear judgment method to obtain the judgment result, which includes linear relationship and nonlinear relationship;

[0010] Data processing: Based on the judgment results, the basic data is integrated and segmented to obtain linear paragraph information and nonlinear paragraph information through processing methods. A multi-level analysis model is established based on the linear paragraph information and nonlinear paragraph information, and a multi-level library is established to store the multi-level analysis model;

[0011] Catalog establishment: a catalog table expressing the corresponding relationship between wear resistance and analysis model is established through the catalog establishment method;

[0012] Optimization update: Perform individual optimizations on analysis models in a multi-level library through update methods;

[0013] The methods for establishing the analysis model include:

[0014] Data acquisition and processing: Integrate linear segment information and nonlinear segment information to obtain target information, intercept single segment information in the target information to obtain target data, and use optimization methods to accurately process the target data and mark the wear resistance to obtain training data;

[0015] Model selection and training: Select a deep learning model as the model matrix, import the training data into the model matrix for model training to obtain the initial model;

[0016] Model adjustment and output: obtain verification information and result information, import the verification information into the initial model to obtain verification results, compare the result information and verification results to obtain comparison results, and adjust and optimize the initial model based on the comparison results to obtain the analysis model.

[0017] Furthermore, the linear judgment method includes: splitting the experimental data to obtain wear-resistant data and parameter data, establishing the correlation between the wear-resistant data and the parameter data, sorting the wear-resistant data based on the wear resistance performance to obtain the sorting result, extracting the parameter data of adjacent wear-resistant data based on the sorting result and the correlation to obtain the first data to be processed and the second data to be processed, adjusting the parameters in the first data to be processed single or multiple times through the adjustment method based on the first data to be processed and the second data to be processed to obtain a reference data set, obtaining the glaze wear resistance under the reference data set to obtain a wear-resistant reference data set, integrating adjacent wear-resistant data and wear-resistant reference data sets to obtain overall data, and judging whether it is a linear relationship based on the change of a single data item in the overall data to obtain a judgment result, and the judgment result includes a linear relationship and a nonlinear relationship.

[0018] Furthermore, the adjustment method includes: obtaining parameter items based on parameter data, presetting adjustment quantities and added values, the adjustment quantities including single-item adjustments and multiple-item adjustments, selecting single parameter items in the first data to be processed in turn based on the single-item adjustment, and adding the parameter items step by step with the added values ​​until they are consistent with the parameter items in the second data to be processed to obtain the first part of data, selecting multiple parameter items in the first data to be processed in turn based on the multiple-item adjustment, and adding the parameter items in turn and incrementally with the added values ​​until they are consistent with the parameter items in the second data to be processed to obtain the second part of data, and integrating the first part of data and the second part of data to obtain a reference data set.

[0019] Furthermore, the processing method includes: extracting adjacent wear-resistant data and judgment results thereof to obtain target information, the target information includes linear segment information and nonlinear segment information, the linear segment information includes overall data whose data changes linearly and a reference data set corresponding to the overall data and adjacent wear-resistant data, the nonlinear segment information includes overall data whose data changes nonlinearly and a reference data set corresponding to the overall data and adjacent wear-resistant data.

[0020] Furthermore, the directory establishment method: extracts boundary wear data of single information in linear paragraph information and nonlinear paragraph information to obtain a boundary data set, splits the boundary data set to obtain a plurality of boundary data, obtains corresponding wear data for establishing the analysis model based on the training data of the analysis model, determines the relationship between the corresponding wear data and the boundary data in the boundary data set to obtain result information, establishes the association relationship between the analysis model and the linear paragraph information and the nonlinear paragraph information based on the result information, and integrates the association relationship, boundary data, analysis model, linear paragraph information and nonlinear paragraph information to obtain a directory table.

[0021] Furthermore, the updating method includes: monitoring user feedback, determining information to be updated based on user feedback, searching for an analysis model corresponding to the information to be updated in a multi-level library based on the information to be updated to obtain the model to be updated, retraining, optimizing and adjusting the establishment method of the model to be updated based on the analysis model based on the information to be updated to obtain an optimized model to be updated, and replacing the model to be updated with the optimized model to be updated in the multi-level library to complete a single optimization of the analysis model.

[0022] Furthermore, the optimization method includes: splitting the target data to obtain a number of sub-data, randomly sorting the sub-data to obtain sorting information, presetting a random number range and a number of selections, generating a target number in the random number range based on the number of selections, extracting the corresponding sub-data based on the target number and the sorting information to obtain the data to be verified, conducting experiments based on the data to be verified to obtain experimental results, presetting a qualified rate, judging whether the experimental results are consistent with the wear-resistant data in the data to be verified and calculating the qualified rate in combination with the number of selections to obtain the target rate, when the target rate is greater than the qualified rate, marking the wear-resistant data in the target data to obtain training data, and when the target rate is less than the qualified rate, integrating the target data, the experimental results and the target rate to obtain feedback information, obtaining feedback address information, and feeding back the feedback information to the user based on the feedback address information.

[0023] The correlation modeling and analysis system between the wear resistance of the glaze and the process parameters uses the above-mentioned correlation modeling and analysis method between the wear resistance of the glaze and the process parameters.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] The modeling and analysis method and system for the correlation between glaze wear resistance and process parameters can facilitate users to obtain corresponding parameter data in the analysis model according to their own wear resistance requirements through the set multi-level analysis model, so that users can understand the correlation between wear resistance and process parameters, and produce glaze wear resistance of corresponding requirements according to the parameter data, which is conducive to improving work efficiency. Through the set update method, a single analysis model can be optimized and updated. Through the unique multi-level model design and the update method, a single model can be optimized and updated, which can reduce the data processing volume when training, updating and using the model to improve the response speed and ensure the efficiency of product production. The parameter data does not represent fixed control parameters, which is conducive to the application of different process conditions.

[0026] At the same time, in the linear judgment method, by adopting experimental data as the data basis, the judgment accuracy of the linear judgment method can be improved. Through the linear judgment method, linear and nonlinear data can be separated to facilitate the establishment of subsequent models. Through the set adjustment method, regular data can be obtained, which can facilitate the linear judgment method to judge whether the data is linear or nonlinear. Through the set processing method and the adjustment method, data for subsequent model training can be added, the data volume of training data can be enriched, and the accuracy of the data can be ensured, which is conducive to improving the accuracy of the model. The data is sorted and processed through the processing method to distinguish different data.

[0027] At the same time, through the directory establishment method, a directory table can be established, and the analysis model and the wear-resistant data range that is compatible with the analysis model can be displayed through the directory table, so that users can find the corresponding analysis model in the directory table according to the required target wear-resistant data. Through the set update method, user feedback is monitored and the analysis model is updated according to user needs. During the update process, a single or multiple analysis models can be updated separately, and the use of other analysis models will not be affected during the update process, which is conducive to ensuring efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a schematic diagram of the overall process structure of the present invention;

[0029] Figure 2 Establish a structural schematic diagram for the separation model of the present invention;

[0030] Figure 3 It is a schematic diagram of the adjacent wear-resistant data structure of the present invention;

[0031] Figure 4 Schematic diagram of a single-item adjustment or multi-item adjustment structure of the present invention, wherein (a) is a schematic diagram of a parameter item structure, (b) is a schematic diagram of a single-item adjustment structure, and (c) is a schematic diagram of a multi-item adjustment structure;

[0032] Figure 5 It is a schematic diagram of the structure of the adjustment method of the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] Glaze is a smooth, ceramic coating applied to the surface of materials such as ceramics, porcelain or glass. It is usually made of glass frit and other additives, and after high-temperature firing, it combines with the substrate to form a dense, smooth and shiny surface. The wear resistance of the glaze refers to the ability of the glaze to resist surface wear when subjected to friction or abrasion. This performance is usually evaluated by indicators such as wear rate, friction coefficient or wear depth. The higher the wear resistance, the more the glaze can maintain its appearance and function in daily use, reducing the frequency of repair or replacement.

[0035] like Figure 1-Figure 5 As shown, the present invention provides a technical solution: a correlation modeling and analysis method between glaze wear resistance and process parameters, the method comprising:

[0036] Model establishment: Establish an analytical model to reflect the correlation between glaze wear resistance and process parameters;

[0037] Data acquisition: Obtain production data and experimental data, integrate production data and experimental data to obtain basic data, which includes wear resistance and process parameters corresponding to wear resistance:

[0038] Linear judgment: Based on the basic data, the relationship between the wear resistance and the change of process parameters is judged by the linear judgment method to obtain the judgment result, which includes linear relationship and nonlinear relationship;

[0039] Data processing: Based on the judgment results, the basic data is integrated and segmented to obtain linear paragraph information and nonlinear paragraph information through processing methods. A multi-level analysis model is established based on the linear paragraph information and nonlinear paragraph information, and a multi-level library is established to store the multi-level analysis model;

[0040] Catalog establishment: a catalog table expressing the corresponding relationship between wear resistance and analysis model is established through the catalog establishment method;

[0041] Optimization update: Perform individual optimizations on analysis models in a multi-level library through update methods;

[0042] It should be noted that, in the process of model establishment, an analysis model is established to reflect the correlation between the wear resistance of the glaze and the process parameters. In subsequent use, the user can import the corresponding wear resistance into the analysis model to obtain the parameter data under the wear resistance. At the same time, the wear resistance can also be obtained vice versa, which is convenient for the staff to conduct test operations before production, so as to find the corresponding parameters and wear resistance that meet the production needs of the staff. That is, the correlation between the wear resistance of the glaze and the process parameters is reflected through the analysis model. In the data acquisition stage, the production data is obtained by acquiring the previous production data, and the experimental data is obtained by retrieval or by the staff through precise experiments. In the linear judgment stage, the basic data is processed by the linear judgment method to judge the relationship between the wear resistance and the process parameters in the basic data, and specifically judge whether the relationship is linear or nonlinear. In the data processing stage, the basic data is integrated through the processing method according to the judgment result. The linear and nonlinear paragraph information corresponding to the judgment results are obtained by merging the segments. The linear and nonlinear paragraph information are used as basic training data to establish a multi-level analysis model. A multi-level library is established to store the multi-level analysis model. The multi-level analysis model can be specifically understood as multiple analysis models corresponding to multiple different wear resistance performance intervals. In the directory establishment stage, a directory table for expressing the relationship between wear resistance and analysis model is established according to the linear and nonlinear paragraph information in combination with the directory establishment method. The corresponding analysis model can be easily found through the directory table to explore the correlation between wear resistance and process parameters. In the optimization and update stage, a single analysis model can be optimized and updated through the set update method. The single model can be optimized and updated through the unique multi-level model design and the update method, which can reduce the data processing volume when training, updating and using the model to improve the response speed and ensure the efficiency of product production.

[0043] In the specific implementation process: first, it is necessary to obtain the production data and experimental data of the previous glazed products to obtain the basic data, which specifically includes the specific wear resistance and the process parameters that cause the specific wear resistance. For example, the process parameters are: firing temperature 1200°C, firing time 6 hours, corresponding wear resistance: wear resistance coefficient is 0.5 m³ / N·km, Mohs hardness 7, process parameters are firing temperature 1300°C, firing time 4 hours, corresponding wear resistance: wear resistance coefficient is 0.3 m³ / N·km, Mohs hardness 8, the linear judgment method is used to judge the relationship of the information in the basic data, and then the basic data is integrated and segmented according to the judgment relationship through the processing method, and the linear and nonlinear information is obtained. An analysis model is established based on the linear and nonlinear information, and then a multi-level library is established to store the analysis model. A catalog table is established through the target establishment method to express the relationship between the wear resistance and the analysis model. For example, when the wear resistance is X, it corresponds to analysis model 1. The analysis model in the multi-level library is optimized and updated through the update method. In actual use, it is first necessary to obtain the user's needs and feedback address information. The user's needs are the corresponding parameters that the user needs to obtain specific wear resistance. The feedback address information is the user's address information, including but not limited to email addresses, mobile phone numbers, etc. According to the user's needs, the corresponding analysis model is retrieved in the catalog table, the analysis model is extracted in the multi-level library, and the wear resistance required by the user is imported into the analysis model to obtain the corresponding production parameters, so that the user can produce according to actual use needs.

[0044] like Figure 2 As shown, the method for establishing the analysis model includes:

[0045] Data acquisition and processing: Integrate linear segment information and nonlinear segment information to obtain target information, intercept single segment information in the target information to obtain target data, and use optimization methods to accurately process the target data and mark the wear resistance to obtain training data;

[0046] Model selection and training: Select a deep learning model as the model matrix, import the training data into the model matrix for model training to obtain the initial model;

[0047] Model adjustment and output: obtain verification information and result information, import the verification information into the initial model to obtain verification results, compare the result information and verification results to obtain comparison results, and adjust and optimize the initial model based on the comparison results to obtain the analysis model.

[0048] It should be noted that in the data acquisition and processing stage, a single paragraph of target information is intercepted as target data, and after optimization, it is used as training data. The expression of the analysis model corresponds to the wear resistance performance range in the intercepted single paragraph of information. The target data is refined and marked with wear resistance processing through the set optimization method to improve the accuracy of the target data, which is conducive to the training of subsequent models. In the model selection and training stage, a suitable deep learning model is selected as the model matrix. The deep learning model includes but is not limited to convolutional neural networks, deep belief networks, etc. The training data is imported into the model for the initial model training. In the model adjustment and output stage, the verification information and result information are the test data used to test the initial model and the corresponding results with the test data. The verification results are obtained by importing the verification information into the initial model. The initial model is adjusted and optimized in combination with the verification results and the corresponding results to obtain the analysis model.

[0049] In the specific implementation process, the training data, for example, the firing temperature is 1250°C, the glazing thickness is 1.0mm, the composition is 50% quartz + 30% feldspar + 20% clay, the wear resistance coefficient is 0.45m³ / N·km, and the hardness value is 700HV. Sample 2: the firing temperature is 1300°C, the glazing thickness is 0.8mm, the composition is 70% quartz + 20% feldspar + 10% iron ore, the wear resistance coefficient is 0.35m³ / N·km, and the hardness value is 720H. Specifically, the wear resistance performance needs to be marked. When selecting the model, the multilayer perceptron can be used as the basic model. The multilayer perceptron is suitable for processing structured data and can better The nonlinear relationship between input features and output is captured accurately. The processed and labeled training data is input into the model matrix. The number of training iterations is set to 1000, and the validation loss value is monitored to prevent overfitting. The initial model is evaluated using the validation set, and the prediction results of the model on the validation data are obtained. The validation results are compared with the actual validation data, and the prediction error of the model is calculated. According to the error analysis of the validation results, the hyperparameters of the model (such as learning rate, number of hidden layer nodes, etc.) are adjusted, feature selection is performed, and the process parameters that have a greater impact on the prediction of wear resistance are retained. The final analysis model is obtained after multiple iterative optimizations, and the verified analysis model is exported.

[0050] like Figure 1 and Figure 3As shown, the linear judgment method includes: splitting the experimental data to obtain wear-resistant data and parameter data, establishing the correlation between the wear-resistant data and the parameter data, sorting the wear-resistant data based on the wear-resistant performance to obtain the sorting result, extracting the parameter data of the adjacent wear-resistant data based on the sorting result and the correlation to obtain the first data to be processed and the second data to be processed, adjusting the parameters in the first data to be processed by the adjustment method based on the first data to be processed and the second data to be processed to obtain a reference data set, obtaining the glaze wear resistance under the reference data set to obtain a wear-resistant reference data set, integrating the adjacent wear-resistant data and the wear-resistant reference data set to obtain the overall data, judging whether it is a linear relationship based on the change of a single data item in the overall data to obtain a judgment result, and the judgment result includes a linear relationship and a nonlinear relationship.

[0051] It should be noted that in the linear judgment method, by using experimental data as the data basis, the judgment accuracy of the linear judgment method can be improved. The wear-resistant data is sorted based on the wear resistance to obtain the sorting result. That is, the wear resistance is sorted from good to bad. After sorting, the parameters of adjacent wear-resistant data are extracted as boundaries, and the parameters in the two boundaries are obtained through adjustment methods for single or multiple adjustments to obtain a reference data set. The wear resistance of each individual data under the reference data set is obtained and integrated to obtain a wear-resistant reference data set. The extracted adjacent wear-resistant data and the wear-resistant reference data set are integrated to obtain the overall data. The judgment result can be obtained by judging whether the change of a single data item in the overall data is a linear relationship.

[0052] In the specific implementation process, it is first necessary to split the experimental data and establish the correlation between the wear resistance data and the parameter data for subsequent search. In the process of sorting the wear resistance data, according to the actual use needs, select a performance indicator for sorting, or sort by combining multiple indicators. For example, when the wear resistance data is sample 1: the wear resistance coefficient is 0.45m³ / N·km, and the hardness value is 700HV, sample 2: the wear resistance coefficient is 0.35m³ / N·km, and the hardness value is 720HV, sample 3: the wear resistance coefficient is 0.50m³ / N·km, and the hardness value is 680HV, sample 4: the wear resistance coefficient is 0.30m³ / N·km, and the hardness value is 740HV, sample 5: The wear coefficient is 0.40m³ / N·km, and the hardness value is 710HV. When a single performance indicator is selected for sorting, when the wear resistance coefficient is selected for sorting, the sorting results are sample 3, sample 1, sample 5, sample 2, and sample 4. If multiple factors need to be considered comprehensively (for example, wear resistance coefficient and hardness value), weighted average or other evaluation methods can be used. Assume that the weight of the wear resistance coefficient is set to 0.7, the weight of the hardness value is set to 0.3, and the hardness value is normalized: weighted score = 0.7×wear resistance coefficient + 0.3×(hardness value-minimum hardness value) / (maximum hardness value-minimum hardness value). The sorting results will display the sample numbers with high and low comprehensive performance scores and their corresponding scores.

[0053] like Figure 4 and Figure 5 As shown, the adjustment method includes: obtaining parameter items based on parameter data, presetting adjustment quantities and added values, the adjustment quantities including single-item adjustment and multiple-item adjustment, based on the single-item adjustment, sequentially selecting single parameter items in the first data to be processed and cooperating with the added values ​​to add the parameter items step by step until they are consistent with the parameter items in the second data to be processed to obtain the first part of data, based on the multiple-item adjustment, sequentially selecting multiple parameter items in the first data to be processed and cooperating with the added values ​​to add the parameter items in sequence and incrementally until they are consistent with the parameter items in the second data to be processed to obtain the second part of data, and integrating the first part of data and the second part of data to obtain a reference data set.

[0054] It should be noted that the parameter item is a single data item in the parameter data, such as firing temperature and glazing thickness. The adjustment quantity and the added value are formulated according to the actual usage. The adjustment quantity can be one or more, namely single-item adjustment and multiple-item adjustment. The added value corresponds to the parameter item, such as 5°C for the firing temperature and 0.5mm for the glazing thickness. Based on the single-item adjustment, the single parameter item in the first data to be processed is selected in turn and the added value is used to add the parameter item step by step until the parameter item is consistent with the parameter item in the second data to be processed to obtain the first part of the data. That is, the added value is added to the numerical value of the parameter item in the first data to be processed, that is, the next addition is carried out on the basis of the previous addition. Through the adjustment method, regular data can be obtained, which can facilitate the judgment of linear or nonlinear results, which is conducive to the subsequent steps. At the same time, by obtaining regular data, the training set used for training the model can be enriched, and the data is obtained through experiments, which is conducive to the accurate training of the model.

[0055] In the specific implementation process, when the parameter data is the firing temperature, firing time and glazing thickness, the parameter item is the corresponding controllable parameter for controlling the firing temperature, firing time and glazing thickness, and the adjustment quantity can be single or multiple. When the adjustment quantity is single, the corresponding controllable parameter of the firing temperature, firing time or glazing thickness is adjusted by adding the added value separately. When the adjustment quantity is multiple, it is two or three items at this time, that is, the two or three controllable parameters are adjusted by adding the added value at the same time, such as Figure 4 As shown in (a), when there are three parameters, the single item adjustment is as follows: Figure 4 As shown in (b), multiple adjustments are Figure 4 As shown in (c).

[0056] like Figure 3 As shown, the processing method includes: extracting adjacent wear-resistant data and judgment results thereof to obtain target information, the target information includes linear segment information and nonlinear segment information, the linear segment information includes overall data whose data changes linearly and a reference data set corresponding to the overall data and adjacent wear-resistant data, the nonlinear segment information includes overall data whose data changes nonlinearly and a reference data set corresponding to the overall data and adjacent wear-resistant data.

[0057] It should be noted that by setting the processing method in conjunction with the adjustment method, you can add data for subsequent model training, enrich the amount of training data, and ensure the accuracy of the data, which is conducive to improving the accuracy of the model. The data is sorted and processed by the processing method to facilitate the distinction between different data. In the specific use process, when the linear judgment method is carried out, the processing method is carried out simultaneously. The processing method is used to collect the data processed in the linear judgment method, and the data is collected and sorted through the processing method.

[0058] like Figure 1 As shown, the directory establishment method is as follows: the boundary wear data of single information in linear paragraph information and nonlinear paragraph information are extracted to obtain a boundary data set, the boundary data set is split to obtain a plurality of boundary data, corresponding wear data for establishing the analysis model are obtained based on the training data of the analysis model, the relationship between the corresponding wear data and the boundary data in the boundary data set is determined to obtain result information, the association relationship between the analysis model and the linear paragraph information and the nonlinear paragraph information is established based on the result information, and the association relationship, boundary data, analysis model, linear paragraph information and nonlinear paragraph information are integrated to obtain a directory table.

[0059] It should be noted that the process of extracting boundary wear data of single information in linear paragraph information and nonlinear paragraph information to obtain boundary data set is to extract single paragraph information in linear paragraph information and nonlinear paragraph information, and the paragraph information is linear paragraph information or nonlinear paragraph information. The boundary data is obtained by extracting the maximum and minimum wear data in the paragraph information. The corresponding wear data for establishing the analysis model is obtained according to the boundary data combined with the training data of the analysis model. An association relationship is established according to the result information, that is, when the corresponding wear data is consistent with the boundary data in the boundary data set, an association relationship is established. Through the catalog establishment method, a catalog table can be established. The analysis model and the wear data range adapted to the analysis model are displayed through the catalog table, which can facilitate users to find the corresponding analysis model in the catalog table according to the required target wear data.

[0060] In the specific implementation process, multiple information in the linear paragraph information and the nonlinear paragraph information are split into single paragraph information, the boundary wear-resistant data in the paragraph information are extracted and integrated to obtain the boundary data set. When training the analysis model, the analysis model is marked with the training data, and the training data is matched with the boundary data in the boundary data set. After the match is successful, the relationship between the analysis model and the boundary data is established, and then the association relationship between the analysis model and the linear paragraph information and the nonlinear paragraph information is established. After completing the matching of all analysis models, the directory table can be established.

[0061] like Figure 1 As shown, the updating method includes: monitoring user feedback, determining information to be updated based on user feedback, searching for an analysis model corresponding to the information to be updated in a multi-level library based on the information to be updated to obtain the model to be updated, retraining, optimizing and adjusting the establishment method of the model to be updated based on the analysis model based on the information to be updated to obtain an optimized model to be updated, and replacing the model to be updated with the optimized model to be updated in the multi-level library to complete a single optimization of the analysis model.

[0062] It should be noted that the determination of whether to update is performed is made by monitoring user feedback. When user feedback indicates that an update is required, the information to be updated is determined based on the user feedback, and the model to be updated and the corresponding training data are determined based on the information to be updated. The model to be updated is retrained, optimized, adjusted and replaced through the training data to complete the update and optimization of the analysis model. Through the set update method, the analysis model can be updated according to user needs, and during the update process, a single or multiple analysis models can be updated separately, and the use of other analysis models will not be affected during the update process, which is conducive to ensuring efficiency.

[0063] During the specific implementation process, a feedback collection interface is added to receive user feedback. When the feedback collection interface collects information that requires updating the analysis model, the corresponding analysis model is determined based on the information, and the analysis model is retrained, optimized, adjusted, and the original analysis model is replaced based on the training data in the information in terms of the method of establishing the analysis model, thereby completing the update and optimization of the analysis model.

[0064] like Figure 2 As shown, the optimization method includes: splitting the target data to obtain a number of sub-data, randomly sorting the sub-data to obtain sorting information, presetting a random number range and a selection number, generating a target number in the random number range based on the selection number, extracting the corresponding sub-data based on the target number and the sorting information to obtain the data to be verified, conducting experiments based on the data to be verified to obtain experimental results, presetting a qualified rate, judging whether the experimental results are consistent with the wear-resistant data in the data to be verified and calculating the qualified rate in combination with the selection number to obtain the target rate, when the target rate is greater than the qualified rate, marking the wear-resistant data in the target data to obtain training data, and when the target rate is less than the qualified rate, integrating the target data, the experimental results and the target rate to obtain feedback information, obtaining feedback address information, and feeding back the feedback information to the user based on the feedback address information.

[0065] It should be noted that the feedback address information is the user's address information, including but not limited to email address, telephone number, etc. The random number range and the number of selections are determined by the user according to actual usage needs. The specific random number range needs to be consistent with the sorting information. The number of selections is the number of times the sub-data is selected. The experimental results are obtained by selecting sub-data and conducting experiments. The qualified rate is determined according to actual usage. The target rate is obtained by combining the consistency between the experimental results and the wear-resistant data in the data to be verified and the number of selections. When the target rate is less than the qualified rate, feedback information is generated and fed back to the user. The user decides whether to conduct model training or conduct experiments to collect training data again to ensure the accuracy of the analysis model.

[0066] In the specific implementation process, when the sorting information is 30, the random number range is also 30, and the number of selections can be 5. Five random numbers are randomly generated through the random number range to obtain the target coordinates. When the target coordinates are 7, 11, 17, 20 and 28, the sub-data corresponding to the target coordinates in the sorting information are extracted to obtain the data to be verified. Experiments are conducted according to the parameter information in the data to be verified and the experimental results are obtained. It is determined whether the experimental results are consistent with the wear-resistant data in the data to be verified, and the target rate is obtained in combination with the number of selections. When there is 1 experimental result in 5 selections that is inconsistent with the wear-resistant data in the data to be verified, the target rate is 80%. When the qualified rate is 100%, the target data, experimental results and target rate are integrated and fed back to the user. When the qualified rate is 60%, the wear-resistant data in the target data is marked to obtain training data.

[0067] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is limited by the attached embodiments and their equivalents.

Claims

1. A modeling and analysis method for the correlation between glaze wear resistance and process parameters, characterized in that: The method comprises: Data acquisition: Obtain production data and experimental data to obtain basic data: Linear judgment: linear and nonlinear relationships are obtained through linear judgment methods based on basic data; Model establishment: Establish an analytical model to reflect the correlation between glaze wear resistance and process parameters; Data processing: The basic data is processed by a processing method to obtain linear paragraph information and nonlinear paragraph information, a multi-level analysis model is established based on the linear paragraph information and the nonlinear paragraph information, and a multi-level library is established to store the multi-level analysis model; Optimization update: Perform individual optimizations on analysis models in a multi-level library through update methods; The methods for establishing the analysis model include: Data acquisition and processing: Integrate linear segment information and nonlinear segment information to obtain target information, intercept single segment information in the target information to obtain target data, and use optimization methods to accurately process the target data and mark the wear resistance to obtain training data; Model selection and training: Select a deep learning model as the model matrix, import the training data into the model matrix for model training to obtain the initial model; Model adjustment and output: obtain verification information and result information, import the verification information into the initial model to obtain verification results, compare the result information and verification results to obtain comparison results, and adjust and optimize the initial model based on the comparison results to obtain the analysis model.

2. The correlation modeling and analysis method between the wear resistance of glaze and process parameters according to claim 1, characterized in that: The linear judgment method includes: splitting experimental data to obtain wear-resistant data and parameter data, establishing the correlation between the wear-resistant data and the parameter data, sorting the wear-resistant data based on the wear-resistant performance to obtain a sorting result, extracting the parameter data of adjacent wear-resistant data based on the sorting result and the correlation to obtain the first data to be processed and the second data to be processed, adjusting the parameters in the first data to be processed by a single or multiple adjustment method based on the first data to be processed and the second data to be processed to obtain a reference data set, obtaining the glaze wear-resistant performance under the reference data set to obtain a wear-resistant reference data set, integrating adjacent wear-resistant data and the wear-resistant reference data set to obtain overall data, judging whether it is a linear relationship based on the change of a single data item in the overall data to obtain a judgment result, and the judgment result includes a linear relationship and a nonlinear relationship.

3. The correlation modeling and analysis method between the wear resistance of glaze and process parameters according to claim 2 is characterized in that: The adjustment method includes: obtaining parameter items based on parameter data, presetting adjustment quantities and added values, the adjustment quantities including single-item adjustment and multiple-item adjustment, selecting single parameter items in the first data to be processed in turn based on the single-item adjustment, and adding the parameter items step by step with the added values ​​until they are consistent with the parameter items in the second data to be processed to obtain a first part of data, selecting multiple parameter items in the first data to be processed in turn based on the multiple-item adjustment, and adding the parameter items in turn and incrementally with the added values ​​until they are consistent with the parameter items in the second data to be processed to obtain a second part of data, and integrating the first part of data and the second part of data to obtain a reference data set.

4. The correlation modeling and analysis method between the wear resistance of glaze and process parameters according to claim 2 is characterized in that: The processing method includes: extracting adjacent wear-resistant data and judgment results thereof to obtain target information, wherein the target information includes linear segment information and nonlinear segment information, wherein the linear segment information includes overall data in which data changes linearly, a reference data set corresponding to the overall data, and adjacent wear-resistant data, and the nonlinear segment information includes overall data in which data changes nonlinearly, a reference data set corresponding to the overall data, and adjacent wear-resistant data.

5. The correlation modeling and analysis method between glaze wear resistance and process parameters according to claim 1, characterized in that: A catalog table expressing the corresponding relationship between wear resistance and analysis model is established through a catalog establishment method, wherein the catalog establishment method comprises: extracting boundary wear resistance data of single information in linear paragraph information and nonlinear paragraph information to obtain a boundary data set, splitting the boundary data set to obtain a plurality of boundary data, obtaining corresponding wear resistance data for establishing the analysis model based on the training data of the analysis model, determining the relationship between the corresponding wear resistance data and the boundary data in the boundary data set to obtain result information, establishing the correlation relationship between the analysis model and the linear paragraph information and the nonlinear paragraph information based on the result information, and integrating the correlation relationship, boundary data, analysis model, linear paragraph information and nonlinear paragraph information to obtain a catalog table.

6. The correlation modeling and analysis method between glaze wear resistance and process parameters according to claim 1, characterized in that: The updating method includes: monitoring user feedback, determining information to be updated based on the user feedback, searching for an analysis model corresponding to the information to be updated in a multi-level library based on the information to be updated to obtain the model to be updated, retraining, optimizing and adjusting the establishment method of the model to be updated based on the analysis model based on the information to be updated to obtain an optimized model to be updated, and replacing the model to be updated with the optimized model to be updated in the multi-level library to complete a single optimization of the analysis model.

7. The correlation modeling and analysis method between glaze wear resistance and process parameters according to claim 1, characterized in that: The optimization method includes: splitting the target data to obtain a number of sub-data, randomly sorting the sub-data to obtain sorting information, presetting a random number range and a selection number, generating a target number in the random number range based on the selection number, extracting the corresponding sub-data based on the target number and the sorting information to obtain the data to be verified, performing experiments based on the data to be verified to obtain experimental results, presetting a qualified rate, judging whether the experimental results are consistent with the wear-resistant data in the data to be verified and calculating the qualified rate in combination with the selection number to obtain the target rate, when the target rate is greater than the qualified rate, marking the wear-resistant data in the target data to obtain training data, and when the target rate is less than the qualified rate, integrating the target data, the experimental results and the target rate to obtain feedback information, obtaining feedback address information, and feeding back the feedback information to the user based on the feedback address information.

8. Modeling and analysis system for the correlation between glaze wear resistance and process parameters, characterized by: The correlation modeling and analysis method between the wear resistance of the glaze and the process parameters as described in any one of claims 1 to 7 is used.

Images