Intelligent design evaluation method for silicon steel product production scheme

Abstract

The application discloses an intelligent design evaluation method for a silicon steel product production scheme, and comprises the following steps: 1, collecting data, and integrating user data, laboratory research and development data and mass production data; 2, constructing user data topics, laboratory research and development data topics and mass production data topics; 3, inputting the type, specification and performance requirement demand information of the silicon steel product, and generating a plurality of production schemes of the silicon steel product according to the demand information, wherein the production schemes comprise an optimized product production scheme and a new product production scheme; 4, multi-dimensionally comprehensively evaluating each production scheme of the silicon steel product generated in the step 3, obtaining a green design index of the silicon steel product production scheme, and recommending the production scheme according to the green design index. The application is based on product design requirements, integrates laboratory research and development data and mass production data, generates and evaluates a product design scheme model, shortens a research and development cycle, and improves research and development efficiency.

Description

Technical Field

[0001] This invention relates to a steel product manufacturing scheme, and more particularly to an intelligent design evaluation method for a silicon steel product manufacturing scheme. Background Technology

[0002] Currently, the research and development of steel products both domestically and internationally mainly relies on expert experience or laboratory trial-and-error simulation experiments. Conventional steel product production schemes involve numerous and complex processes from user requirements to the final transformation into compliant steel products. These include: requirements analysis, failure mode analysis, research scheme formulation, laboratory experiments on chemical composition and process parameters, product trial production output, large-scale industrial production trials, process re-optimization, and user certification. Existing technology design models not only have long development cycles and high product development costs, but also heavily rely on expert subjective experience and theoretical knowledge. Furthermore, during the on-site trial production process from laboratory product trial production schemes, the differences between laboratory equipment and steel mill industrial production line processes introduce significant uncertainties into the laboratory product schemes. Certain components or process parameters may require repeated laboratory verification, further extending the development cycle. In addition, key indicators such as yield, energy consumption, and cost at each production stage cannot be accurately assessed under existing models.

[0003] Electrical steel, also known as silicon steel, is a type of steel product with excellent soft magnetic properties. It is mainly used to manufacture the cores of various motors and transformers, and is an important soft magnetic metallic material in the power and motor industries. Its manufacturing process is complex, especially for grain-oriented silicon steel products. The entire production process includes processes such as molten iron pretreatment, RH refining, continuous casting, hot rolling, normalizing annealing, pickling, cold rolling, decarburizing annealing, MgO coating, high-temperature annealing, leveling and stretching annealing, and coating with an insulating layer. Higher grades also include processes such as nitriding and laser marking. The steelmaking composition and inclusion elements are strictly controlled, and there are many process parameters and influencing factors throughout the process. These involve metal solidification, rolling deformation, recrystallization nucleation and growth control, inhibitor solid solution precipitation control, secondary recrystallization, preferred grain orientation, steel plate decarburization, desulfurization, nitriding and denitrification control, and surface oxide layer control, etc., making it a very complex process. When developing new silicon steel products, all of the above-mentioned influencing factors need to be considered and designed and demonstrated in the experimental scheme. If only expert experience or laboratory trial and error experiments are relied upon, the development cycle will be very long, and the accuracy, rationality and manufacturability of the product design scheme cannot be guaranteed. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent design and evaluation method for silicon steel product manufacturing schemes. Based on product design requirements, it integrates laboratory R&D data and large-scale production data to generate and evaluate product design scheme models, thereby shortening the R&D cycle and improving R&D efficiency.

[0005] This invention is implemented as follows:

[0006] A smart design evaluation method for silicon steel product manufacturing schemes includes the following steps:

[0007] Step 1: Collect data and integrate user data, laboratory R&D data, and large-scale production data;

[0008] Step 2: Construct user data themes, laboratory R&D data themes, and large-scale production data themes;

[0009] Step 3: Input the requirements for the type, specifications, and performance of silicon steel products, and generate several production plans for silicon steel products based on the requirements. The production plans include production plans for optimized products and production plans for new products.

[0010] Step 4: Conduct a multi-dimensional comprehensive evaluation of the production plan for each silicon steel product generated in Step 3 to obtain the green design index of the silicon steel product production plan, and recommend a production plan based on the green design index.

[0011] The user data includes users' performance requirements for various types of silicon steel products, user ID, inquiry ID, and basic user information;

[0012] The laboratory research and development data mentioned above includes chemical composition, research and development process parameters, various performance indicators, and microstructure analysis results throughout the entire experimental process of various types of silicon steel products;

[0013] The aforementioned large-scale production data refers to the production data of each unit in the entire large-scale production process of various types of silicon steel products, including chemical composition, steel tapping marks, production material tracking information, actual production process parameters, various performance data, inspection and testing data, surface defect data, quality judgment data, inlet / outlet coil weight, yield rate, inlet coil weight allocation, energy consumption per ton of steel, and cost per ton of steel. Among them, the actual production process parameters include low-frequency data collected by steel coil and high-frequency data collected by distance or time interval. Each steel coil has one low-frequency data and multiple high-frequency data.

[0014] Step 2 includes:

[0015] Step 2.1: Construct the user demand data theme;

[0016] Step 2.2: Construct laboratory R&D data themes for silicon steel products;

[0017] Step 2.3: Construct a large-scale production data theme for silicon steel products.

[0018] Step 2.1 includes: based on the user ID and inquiry ID in the user data, linking the basic information of a user with the performance requirements of the user for silicon steel products to form a user demand data theme;

[0019] Step 2.2 includes:

[0020] Step 2.2.1: Configure laboratory R&D data anomaly rules for the chemical composition, R&D process parameters, and performance indicators of silicon steel products;

[0021] Step 2.2.2: Preprocess the laboratory R&D data based on the anomaly rules of the laboratory R&D data;

[0022] Step 2.2.3: Based on the entire experimental process, define the entry sample number and exit sample number for each experimental procedure. Using the logic that the exit sample number of the previous experimental procedure equals the entry sample number of the next experimental procedure, connect the entire experimental process data of silicon steel products to form the laboratory R&D data theme.

[0023] Step 2.3 includes:

[0024] Step 2.3.1: Configure large-scale production data anomaly rules for the chemical composition, actual production process parameters, and performance data of silicon steel products;

[0025] Step 2.3.2: Preprocess the large-scale production data based on the anomaly rules of the large-scale production data;

[0026] Step 2.3.4: Calculate the high-frequency data and performance characteristic values ​​of the silicon steel coils;

[0027] Step 2.3.5: Based on the entire large-scale production process, define the exit roll number and inlet roll number of each unit. Using the logic that the exit roll number of the previous unit equals the inlet roll number of the next unit, connect the large-scale production process data of silicon steel products to form a large-scale production data theme.

[0028] Step 3 includes:

[0029] Step 3.1: Filter the laboratory R&D data theme and the large-scale production data theme according to the demand information, and select two types of relevant data that meet the demand information; among them, the performance indicators in the laboratory R&D data theme are multiplied by an adjustment factor before filtering;

[0030] Step 3.2: Based on the performance requirements in the demand information, normalize and standardize the data for each type of relevant data;

[0031] Step 3.3: For each type of relevant data, calculate the first performance similarity between the performance requirements and the normalized and standardized performance data;

[0032] Step 3.4: Set a performance similarity threshold. If the first performance similarity is greater than or equal to the performance similarity threshold, then mark the silicon steel product corresponding to the first performance similarity as an optimized product and proceed to step 3.5. If the first performance similarity is less than the performance similarity threshold, then mark the silicon steel product corresponding to the first performance similarity as a new product and proceed to step 3.6.

[0033] Step 3.5: Generate an optimized production plan for the product: Combine the actual production process parameters corresponding to the first performance similarity with the control precision of each unit in mass production, and divide the actual production process parameters.

[0034] Step 3.6: Based on performance requirements, establish a performance index prediction model and verify the optimal production plan model for the new product.

[0035] Step 3.6 includes:

[0036] Step 3.6.1: Select historical data from the large production data. The types and specifications of silicon steel products in this historical data are the same as those of the input silicon steel products, and the time range of the large production data is the most recent year.

[0037] Step 3.6.2: Establish several performance index prediction models; the input of each performance index prediction model is the key chemical components and key process parameters selected by the R&D personnel, and the output is the performance index. The optimal performance prediction model among several performance index prediction models is determined through cross-validation.

[0038] Step 3.6.3: Calculate the average value of the actual performance of the chemical composition and production process parameters in the large-scale production data selected in Step 3.6.1, and use the average value of the actual performance of the chemical composition and production process parameters as the benchmark point of the scheme;

[0039] Step 3.6.4: Perform a gridded search on the performance index prediction model starting from the baseline of the scheme.

[0040] In step 4, the evaluation dimensions include performance similarity, yield, energy consumption per ton of steel, and cost per ton of steel.

[0041] For each of the aforementioned optimized product production plans, the multi-dimensional comprehensive evaluation method is as follows:

[0042] Step 4.11: Calculate the second performance similarity between the performance data of the optimized product's production plan and the actual performance data of mass production. If the second performance similarity is fully satisfied, that is, the performance data of the optimized product's production plan and the actual performance data of mass production are completely matched, then the first performance score is recorded as 100 points, and proceed to step 4.13. If the second performance similarity is not fully satisfied, then proceed to step 4.12.

[0043] Step 4.12: Calculate the similarity between the magnetic induction and iron loss of the optimized product's production plan and the actual magnetic induction and iron loss in mass production, and perform performance evaluation based on weights to obtain the third performance similarity S3. The evaluation formula is: S3 = λ 11 *S 磁感1 +λ 21 *S 铁损1 The first performance score is recorded as S3;

[0044] Among them, S 磁感1 To optimize the similarity between the magnetic properties of the product's manufacturing performance data and the actual magnetic properties of its mass production performance data, λ 11 For S 磁感1 The weights of S; 铁损1 To optimize the similarity between the magnetic properties of the product's manufacturing performance data and the magnetic properties of the actual performance data in mass production, λ 21 For S 铁损1 The weights; λ 11 +λ 21 =1;

[0045] Step 4.13: Based on the weighted evaluation of the production scheme model of the product, the first yield rate of each unit is evaluated and optimized to obtain the first yield rate index;

[0046] The formula for calculating the first yield of each unit is:

[0047] First yield rate η1 = (export weight / weight of imported coil) * 100;

[0048] Wherein, export weight = ∑(weight of each steel coil exported), and the weight of each steel coil allocated to the inlet coil = ∑(weight of each steel coil allocated to the inlet coil);

[0049] The formula for calculating the first yield rate index is: First yield rate index = ∑(λ3*η1), where λ3 is the weight of the first yield rate corresponding to each unit, and ∑λ3 = 1. The score of the first yield rate is recorded as the first yield rate index.

[0050] Step 4.14: Optimize the energy consumption of the production scheme model based on the energy consumption of steel coil production unit. When the energy consumption is the lowest, the first energy consumption score is recorded as 100 points; when the energy consumption is the highest, the first energy consumption score is recorded as 0 points.

[0051] Step 4.15: Optimize the production plan model based on the cost per ton of steel produced from steel coils. When the cost per ton of steel is the lowest, the first cost score is recorded as 100 points; when the cost per ton of steel is the highest, the first cost score is recorded as 0 points.

[0052] Step 4.16: Calculate the green design index of the optimized product based on weights. The calculation formula is as follows:

[0053] Optimize the product's green design index = λ 41 *First performance score +λ 42 *First yield score +λ 43 *First Energy Efficiency Rating +λ 44 *First cost rating;

[0054] Where, λ 41 λ is the weight for the first performance score. 42 λ is the weight for the first yield rate score. 43 λ is the weight of the first energy consumption score. 44 Let λ be the weight of the first cost score, and λ be the weight of the first cost score. 41 +λ 42 +λ 43 +λ 44 =1.

[0055] For each of the aforementioned new product production plans, the multi-dimensional comprehensive evaluation method is as follows:

[0056] Step 4.21: Calculate the fourth performance similarity between the performance data of the production plan of the new product and the actual performance data of mass production. If it is completely satisfied, that is, the performance data of the production plan of the new product and the actual performance data of mass production are completely matched, then the second performance score is recorded as 100 points, and proceed to step 4.23. If it is not completely satisfied, then execute step 4.22.

[0057] Step 4.22: Calculate the similarity between the magnetic induction and iron loss of the production plan for the new product and the actual magnetic induction and iron loss in mass production, and perform performance evaluation based on weights to obtain the fifth performance similarity S5. The evaluation formula is: S5 = λ 12 *S 磁感2 +λ 22 *S 铁损2 The second performance rating is recorded as S5.

[0058] Among them, S 磁感2 λ represents the similarity between the magnetic properties of the production plan performance data and the actual magnetic properties of the mass production performance data for a new product. 12 For S 磁感2 The weights of S; 铁损2 λ represents the similarity between the magnetic properties of the production plan performance data and the actual magnetic properties of the mass production performance data for a new product. 22 For S 铁损2 The weights; λ 12 +λ 22 =1;

[0059] Step 4.23: Evaluate the second yield rate of each unit in the production plan of the new product based on the weight, and obtain the second yield rate index;

[0060] The formula for calculating the second yield rate of each unit is:

[0061] Second yield rate η2 = (export weight / imported coil allocated weight) * 100;

[0062] Wherein, export weight = ∑(weight of each steel coil exported), and the weight of each steel coil allocated to the inlet coil = ∑(weight of each steel coil allocated to the inlet coil);

[0063] The formula for calculating the second yield rate index is: Second yield rate index = ∑(λ5*η2), and ∑λ5 = 1. The score of the second yield rate is recorded as the second yield rate index.

[0064] Step 4.24: Evaluate the energy consumption of the production plan for the new product based on the energy consumption per unit of steel coil production. When the energy consumption is the lowest, the second energy consumption score is recorded as 100 points; when the energy consumption is the highest, the second energy consumption score is recorded as 0 points.

[0065] Step 4.25: Based on the cost per ton of steel produced from steel coils, evaluate and optimize the production plan for the product. When the cost per ton of steel is the lowest, the second cost score is recorded as 100 points; when the cost per ton of steel is the highest, the second cost score is recorded as 0 points.

[0066] Step 4.26: Calculate the green design index of the new product based on weights. The calculation formula is as follows:

[0067] Green design index of new products = λ 61 *Second performance score +λ 62 *Second yield rate score +λ 63 *Second energy efficiency rating +λ 64 *Second cost rating;

[0068] Where, λ 61 λ is the weight for the second performance score. 62 λ is the weight for the second yield rate score. 63 λ is the weight of the second energy consumption score. 64 λ is the weight of the second cost score, and λ 61 +λ 62 +λ 63 +λ 64 =1;

[0069] Step 4.27: Calculate the deviation coefficient P = P1 + P2 between the new product production plan and the most similar steel grade;

[0070] Wherein, P1 is the first deviation coefficient between the chemical composition of the new product and the chemical composition of the most similar steel grade in the production plan, and the calculation formula is:

[0071]

[0072] Where, λi Assigning the weights to each chemical component, x represents the design values ​​for each chemical component in the production plan of the new product. ik For the actual values ​​of each chemical composition of the most similar steel grade, N i The number of records for the most similar steel grade, where i is the number of chemical compositions;

[0073] P2 is the second deviation coefficient between the process parameters in the new product production plan and the process parameters of the most similar steel grade. The calculation formula is:

[0074]

[0075] Where, β i Assigning weights to each process parameter. g represents the design values ​​of various process parameters in the production plan for the new product. ik N represents the actual values ​​of each process parameter for the most similar steel grade. i The number of records for the most similar steel grade, where i is the number of process parameters;

[0076] Step 4.28: Correct the green design index of the new product using the deviation coefficient P to obtain the corrected green design index of the new product. The correction calculation formula is as follows:

[0077] The corrected green design index of a new product = the green design index of the new product - P.

[0078] Compared with the prior art, the present invention has the following beneficial effects:

[0079] 1. Because this invention integrates data from laboratory research and development and large-scale production, it can design solutions for key chemical components and process parameters throughout the entire production process of silicon steel products, and conduct multi-dimensional comprehensive evaluation of performance indicators. It can trace and adjust data such as any steel coil, any process parameter, and any performance indicator throughout the entire process, which is beneficial to assisting in the optimization of product and new product research and development design.

[0080] 2. Based on the performance requirements of users and R&D, and combined with the large amount of data accumulated from laboratory R&D and mass production, this invention can provide more targeted component product design scheme models, greatly reducing the number of trial and error attempts and the reliance on expert experience, shortening the R&D cycle, and significantly improving R&D efficiency. It has good adaptability and high efficiency for the R&D of products with many and complex production processes, such as silicon steel.

[0081] Based on product design requirements, this invention integrates experimental R&D data and large-scale production data using big data technology to generate product design scheme models and conduct multi-dimensional comprehensive evaluations. It recommends the optimal solution to assist in optimizing products and rapid on-site prototyping of new products, reducing the number of trial and error attempts, shortening the R&D cycle, and significantly improving R&D efficiency. It is especially suitable for the R&D of products with complex production processes, such as silicon steel. Attached Figure Description

[0082] Figure 1 This is a flowchart of the intelligent design evaluation method for silicon steel product production scheme of the present invention. Detailed Implementation

[0083] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0084] Please see the appendix Figure 1 A smart design evaluation method for silicon steel product manufacturing schemes includes the following steps:

[0085] Step 1: Collect data and integrate user data, laboratory R&D data, and large-scale production data.

[0086] The user data includes user performance requirements for various types of silicon steel products (such as iron loss, magnetic induction, and other electromagnetic properties), user ID, inquiry ID, and basic user information.

[0087] The aforementioned laboratory R&D data includes chemical composition, R&D process parameters, various performance indicators (such as iron loss, magnetic induction, and other electromagnetic properties), and microstructure analysis results throughout the entire experimental process for various types of silicon steel products. The entire experimental process includes laboratory steelmaking, laboratory hot rolling, laboratory normalizing, laboratory cold rolling, laboratory continuous annealing, laboratory decarburizing annealing, laboratory magnesium oxide coating, laboratory high-temperature annealing, and laboratory coating processes. Laboratory R&D data can be obtained from the silicon steel product R&D management system, and all data in the laboratory R&D data has been accumulated for at least one year to ensure the accuracy of the production plan. Microstructure analysis results refer to the metallographic structure of steel plates corresponding to each process in the manufacturing of oriented silicon steel products, such as normalized plates produced in the normalizing process, decarburized plates corresponding to the decarburizing annealing process, and high-temperature annealed plates corresponding to the high-temperature annealing process. Metallographic microscopic analysis is performed on these samples to obtain microstructure images, grain sizes, and other data.

[0088] The aforementioned large-scale production data comprises production data from each unit throughout the entire large-scale production process of various types of silicon steel products. This includes chemical composition (e.g., elements C, Si, S, etc.), tapping marks, production material tracking information (e.g., the corresponding unit number, coil entry number, and coil exit number during silicon steel coil production), actual production process parameters (e.g., tapping temperature, maximum normalizing temperature, etc., including low-frequency data collected by coil and high-frequency data collected by distance or time interval; each coil has only one low-frequency data point and multiple high-frequency data points), various performance data (e.g., iron loss, magnetic induction, and other electromagnetic properties), inspection and testing data, surface defect data, quality judgment data, coil weight at entry / exit, yield rate, coil weight allocation at entry, energy consumption per ton of steel, and cost per ton of steel. This large-scale production data can be obtained from the process control system and manufacturing execution system of steel production. All data points in this large-scale production data have been accumulated for at least one year to ensure the accuracy of the constructed solution model.

[0089] Step 2: Construct user data themes, laboratory R&D data themes, and large-scale production data themes.

[0090] Step 2.1: Construct a user demand data topic. The specific operation is as follows: Based on the user ID and inquiry ID in the user data, link the basic information of a user with the performance requirements of the user for silicon steel products to form a user demand data topic.

[0091] Step 2.2: Constructing a laboratory R&D data theme for silicon steel products, the specific steps of which are:

[0092] Step 2.2.1: Configure laboratory R&D data anomaly rules for the chemical composition, R&D process parameters, and performance indicators of silicon steel products. The reasonable data ranges for chemical composition, R&D process parameters, and performance indicators can be determined based on the specific silicon steel product.

[0093] Step 2.2.2: Based on the rules for abnormal laboratory R&D data, preprocess the laboratory R&D data. If the chemical composition, R&D process parameters, or performance indicators exceed their reasonable data range, the laboratory R&D data is judged to be abnormal, and the R&D personnel are notified that there is abnormal data, so that the R&D personnel can handle the abnormal data accordingly. Otherwise, the laboratory R&D data is judged to be normal.

[0094] Step 2.2.3: Based on the entire experimental process, according to the steelmaking date and heat number, the slab opening position of the silicon steel product, and the hot rolling and cold rolling coiling information, define the entry sample number and exit sample number for each experimental process. Using the logic that the exit sample number of the previous experimental process equals the entry sample number of the next experimental process, the entire experimental process data of the silicon steel product is linked together to form the laboratory R&D data theme, which facilitates the traceability of the entire experimental process from steelmaking to cold rolling.

[0095] Step 2.3: Construct a large-scale production data theme for silicon steel products. The specific steps are as follows:

[0096] Step 2.3.1: Configure large-scale production data anomaly rules for the chemical composition, actual production process parameters, and performance data of silicon steel products. The reasonable data ranges for chemical composition, actual production process parameters, and performance data can be determined based on the specific silicon steel product.

[0097] Step 2.3.2: Based on the rules for anomalies in large-scale production data, preprocess the large-scale production data. If the actual chemical composition, production process parameters, and performance data exceed their reasonable range, they are judged as abnormal large-scale production data, and the R&D personnel are notified that there is abnormal data, so that the R&D personnel can handle the abnormal data accordingly. Otherwise, the large-scale production data is judged as normal.

[0098] Step 2.3.4: Calculate the average, maximum, minimum and other characteristic values ​​of the high-frequency data and performance data of the silicon steel coil.

[0099] Step 2.3.5: Based on the entire large-scale production process, define the exit roll number and inlet roll number of each unit. Using the logic that the exit roll number of the previous unit equals the inlet roll number of the next unit, connect the large-scale production process data of silicon steel products to form a large-scale production data theme.

[0100] The large-scale production process data includes actual production process parameters, inspection and testing data, surface defect data, quality judgment data, yield, energy consumption per ton of steel, and cost per ton of steel for each unit. This allows for the traceability of the entire large-scale production process of any steel coil, starting from any process, and includes all data such as actual production process parameters and inspection and testing results for each unit.

[0101] In the aforementioned user demand data theme, laboratory R&D data theme, and large-scale production data theme, the same data fields correspond one-to-one. For example, the user demand data theme, laboratory R&D data theme, and large-scale production data theme all include the same data field "performance". Arranging the data field "performance" in the user demand data theme, laboratory R&D data theme, and large-scale production data theme in a corresponding manner provides a reliable data foundation for subsequent data filtering and selection processing steps in the production plan, thereby improving data processing efficiency and accuracy.

[0102] Step 3: Input the required information such as the type, specifications and performance of silicon steel products, and generate several production plans for silicon steel products based on the required information. The production plans include production plans for optimized products and production plans for new products.

[0103] Step 3.1: Filter laboratory R&D data topics and mass production data topics separately based on the demand information, and select the two types of relevant data that meet the demand information: laboratory R&D related data and mass production related data. Specifically, when screening laboratory R&D data topics, because there are differences between mass production finished product testing standards and laboratory sample performance testing standards, the performance indicators of the laboratory R&D data topics need to be multiplied by an adjustment factor (e.g., 1.2) before participating in the screening. The adjustment factor is set by the R&D personnel based on a comparison of performance indicators of historical mass production data and R&D data under the same process settings. The demand information can be determined based on the user's performance requirements for silicon steel products, or it can be determined based on the R&D personnel's R&D needs for new silicon steel products.

[0104] Step 3.2: Based on the performance requirements in the demand information, perform data processing on the two types of relevant data selected in Step 3.1. Data processing includes normalization and standardization to eliminate the influence of different performance indicators, different units or outliers, and improve the accuracy of production plan generation.

[0105] The normalization formula is: X' = (X - MIN) / (MAX - MIN);

[0106] The standardized formula is: X'=(X-μ) / σ.

[0107] Where X represents the performance requirements of silicon steel products.

[0108] When normalizing laboratory R&D-related data, MIN represents the minimum value of the performance index of the laboratory R&D-related data, MAX represents the maximum value of the performance index of the laboratory R&D-related data, μ represents the average value of the performance index of the laboratory R&D-related data, and σ represents the standard deviation of the main performance index of the laboratory R&D-related data.

[0109] When normalizing large-scale production-related data, MIN represents the minimum value of the performance data, MAX represents the maximum value of the performance data, μ represents the average value of the performance indicators, and σ represents the standard deviation of the performance indicators.

[0110] Normalization and standardization are common data processing methods, and will not be elaborated on here.

[0111] Step 3.3: For each type of relevant data, calculate the first performance similarity between the performance requirements and the normalized and standardized performance data.

[0112] Preferably, the first performance similarity S1 can be calculated using the Euclidean distance similarity algorithm. The calculation formula is: S1 = (1 / (1+DIST))*100, where DIST is the distance between the performance requirement index and the normalized and standardized performance data. The performance data is calculated using the average, maximum, or minimum values ​​of multiple sampling points according to its characteristics. The Euclidean distance similarity algorithm is a conventional technique for data processing and will not be elaborated upon here.

[0113] Step 3.4: Set a performance similarity threshold based on the user's tolerance for the performance deviation range, such as 80%. If the first performance similarity is greater than or equal to the performance similarity threshold, then mark the silicon steel product corresponding to the first performance similarity as an optimized product and proceed to step 3.5. If the first performance similarity is less than the performance similarity threshold, then mark the silicon steel product corresponding to the first performance similarity as a new product and proceed to step 3.6.

[0114] Step 3.5: Generate an optimized production plan for the product: Based on each type of relevant data, combined with the chemical composition, production process parameter performance corresponding to the first performance similarity, and the control precision of each unit in mass production, the R&D personnel select key chemical components (such as element C, element Si, etc.) and key process parameters (such as the maximum normalization temperature, etc.). According to the control precision of each key chemical component and key process parameter, the chemical composition and production process parameter performance are divided one by one. Among them, the production process parameter performance includes the performance of low-frequency process parameters and the average, maximum, minimum, and other characteristic values ​​of high-frequency data. The division ranges of each chemical component and production process parameter performance are combined to generate an optimized production plan for the product.

[0115] Step 3.6: Based on performance requirements, establish a performance index prediction model and verify the optimal production plan for the new product.

[0116] Step 3.6.1: Select historical data from the large production data. The types and specifications of silicon steel products in this historical data are the same as those of the input silicon steel products, and the time range of the large production data is the most recent year, that is, the time range from the current time point to the previous year.

[0117] Step 3.6.2: Establish several performance index prediction models using multi-output XGBoost regression algorithm and neural network algorithm, preferably 3-5 models; the input of each performance index prediction model is the key chemical components and key process parameters selected by the R&D personnel, and the output is the performance index (such as iron loss, magnetic induction and other electromagnetic properties). The optimal performance index prediction model among several performance index prediction models is determined by minimizing the mean absolute error (MAE) of cross-validation.

[0118] Step 3.6.3: Calculate the average value of the actual performance of chemical composition and production process parameters in the large-scale production data selected in Step 3.6.1, and use the average value of the actual performance of chemical composition and production process parameters as the benchmark point of the scheme.

[0119] Step 3.6.4: Starting from the baseline of the proposed scheme, perform a gridded search on the performance index prediction model. This involves generating several schemes based on chemical composition and production process parameter ranges through the gridded search. Then, based on the performance index prediction model from 3.6.2, predict the performance indicators of these schemes to form several production schemes. The control precision of the gridded search can be determined according to actual needs; for example, a typical control precision of ±0.0025 can be used.

[0120] Step 4: Conduct a multi-dimensional comprehensive evaluation of the production plan for each silicon steel product generated in Step 3 to obtain the green design index of the silicon steel product production plan model, and recommend the production plan based on the green design index, that is, the production plan with the highest green design index is the recommended plan.

[0121] Preferably, the evaluation dimensions include performance similarity, yield, energy consumption per ton of steel, and cost per ton of steel.

[0122] For each of the aforementioned optimized product production plans, the multi-dimensional comprehensive evaluation method is as follows:

[0123] Step 4.11: Calculate the second performance similarity S2 between the performance data of the optimized product's production plan and the actual performance data of mass production. The actual performance data of mass production corresponds to the design range of each chemical component and production process in the optimized product's production plan. The second performance similarity S2 is calculated using the Euclidean distance similarity algorithm in Step 3.3. The calculation formula is: S2 = (1 / (1+DIST))*100, where DIST is the distance between the performance data of the optimized product's production plan and the actual performance data of mass production. The performance data is calculated using the average, maximum, or minimum values ​​of multiple sampling points according to its characteristics. The Euclidean distance similarity algorithm is a conventional data processing technique and will not be elaborated upon here. If the performance data is completely satisfied (i.e., the performance data of the production plan perfectly matches the actual performance data of mass production), the first performance score is recorded as 100 points, and the process proceeds to Step 4.13. If the performance data is not completely satisfied, proceed to Step 4.12.

[0124] Step 4.12: Calculate the similarity between the magnetic induction and iron loss of the optimized product's production plan and the actual magnetic induction and iron loss in mass production, and perform performance evaluation based on weights to obtain the third performance similarity S3. The evaluation formula is: S3 = λ 11 *S 磁感1 +λ 21 *S 铁损1The first performance score is recorded as S3.

[0125] Among them, S 磁感1 To optimize the similarity between the magnetic properties of the product's manufacturing performance data and the magnetic properties of the actual performance data from mass production, S 磁感1 The Euclidean distance similarity algorithm described in step 3.3 can be used for calculation, which will not be elaborated here; λ 11 For S 磁感1 The weight.

[0126] S 铁损1 To optimize the similarity between the magnetic properties of the product's manufacturing performance data and the magnetic properties of the actual performance data from mass production, S 铁损1 The Euclidean distance similarity algorithm described in step 3.3 can be used for calculation, which will not be elaborated here; λ 21 For S 铁损1 The weight, λ 11 +λ 21 =1.

[0127] Step 4.13: Based on the weighted evaluation, optimize the first yield rate of each unit in the production plan of the product to obtain the first yield rate index.

[0128] The formula for calculating the first yield of each unit is:

[0129] First yield rate η1 = (export weight / inbound roll allocated weight) * 100, where export weight = ∑(weight of each roll exported), and inbound roll allocated weight = ∑(weight of each roll imported).

[0130] The formula for calculating the first yield rate index is: First yield rate index = ∑(λ3*η1), where λ3 is the weight of the first yield rate corresponding to each unit, and ∑λ3 = 1. The weight allocation of λ3 can be appropriately adjusted by the R&D personnel according to the importance of the unit. The first yield rate score is recorded as the first yield rate index.

[0131] Step 4.14: Optimize the energy consumption per ton of steel in the production plan based on the unit energy consumption of steel coil production. The evaluation formula is: N'=((N-Nmin) / (Nmax-Nmin))*100. That is, when the energy consumption N is the lowest, the first energy consumption score is recorded as 100 points. When the energy consumption N is the highest, the first energy consumption score is recorded as 0 points. The rest are calculated according to the formula. The first energy consumption score is between 0 and 100.

[0132] Where N is the energy consumption per unit of steel coil production, N=∑(Ne) / n, Ne is the energy consumption per ton of steel for each steel coil, n is the number of steel coils, Nmin is the minimum energy consumption per unit of steel coil production, and Nmax is the minimum energy consumption per unit of steel coil production.

[0133] Step 4.15: Optimize the production plan model based on the cost per ton of steel produced from steel coils. The evaluation formula is: C' = ((C-Cmin) / (Cmax-Cmin))*100. When the cost per ton of steel C is the lowest, the first cost score is recorded as 100 points. When the cost per ton of steel C is the highest, the first cost score is recorded as 0 points. The rest are calculated according to the formula. The first cost score is between 0 and 100.

[0134] Where C is the cost per ton of steel produced from a steel coil, C=∑(Ce) / n, Ce is the cost per ton of steel for each steel coil, n is the number of steel coils, Cmin is the minimum cost per ton of steel produced from a steel coil, and Cmax is the maximum cost per ton of steel produced from a steel coil.

[0135] Step 4.16: Calculate the green design index of the optimized product based on weights. The calculation formula is as follows:

[0136] Optimize the product's green design index = λ 41 *First performance score +λ 42 *First yield score +λ 43 *First Energy Efficiency Rating +λ 44 *First cost rating.

[0137] Where, λ 41 λ is the weight for the first performance score. 42 λ is the weight for the first yield rate score. 43 λ is the weight of the first energy consumption score. 44 Let λ be the weight of the first cost score, and λ be the weight of the first cost score. 41 +λ 42 +λ 43 +λ 44 =1.

[0138] For each of the aforementioned new product production plans, the multi-dimensional comprehensive evaluation method is as follows:

[0139] Step 4.21: Calculate the fourth performance similarity S4 between the performance data of the production plan for the new product and the actual performance data of mass production. The actual performance data of mass production corresponds to the design range of each chemical component and production process in the production plan for the new product. The fourth performance similarity S4 can be calculated using the Euclidean distance similarity algorithm in Step 3.3, which will not be elaborated here. If the performance data of the production plan for the new product and the actual performance data of mass production are completely matched, the second performance score is recorded as 100 points (this situation may only occur when the production plan for the new product is generated from R&D data), and proceed to Step 4.23. If the performance data is not completely matched, proceed to Step 4.22.

[0140] Step 4.22: Calculate the similarity between the magnetic induction and iron loss of the production scheme model of the new product and the actual magnetic induction and iron loss in mass production, and perform performance evaluation based on weights to obtain the fifth performance similarity S5. The evaluation formula is: S5 = λ 12 *S 磁感2 +λ 22 *S 铁损2 The second performance score is recorded as S5.

[0141] Among them, S 磁感2 S is the similarity between the magnetic properties of the production plan performance data and the magnetic properties of the actual performance data in mass production of the new product. 磁感2 The Euclidean distance similarity algorithm described in step 3.3 can be used for calculation, which will not be elaborated here; λ 12 For S 磁感2 The weight.

[0142] S 铁损2 S is the similarity between the magnetic properties of the production plan performance data and the magnetic properties of the actual performance data in mass production of the new product. 铁损2 The Euclidean distance similarity algorithm described in step 3.3 can be used for calculation, which will not be elaborated here; λ 22 For S 铁损2 The weight, λ 12 +λ 22 =1.

[0143] Step 3.23: Evaluate the second yield rate of each unit in the production plan of the new product based on the weight, and obtain the second yield rate index.

[0144] The formula for calculating the second yield rate of each unit is:

[0145] Second yield rate η2 = (export weight / inbound roll allocated weight) * 100, where export weight = ∑(weight of each roll exported), and inbound roll allocated weight = ∑(weight of each roll imported).

[0146] The formula for calculating the second yield rate index is: Second yield rate index = ∑(λ5*η2), where λ5 is the weight of the second yield rate η2 of each unit, and ∑λ5 = 1. The R&D personnel can adjust λ5 appropriately according to the importance of the unit, and the second yield rate score is recorded as the second yield rate index.

[0147] Step 4.24: Evaluate the energy consumption of the production plan for the new product based on the energy consumption of the steel coil production unit. The evaluation formula is: N'=((N-Nmin) / (Nmax-Nmin))*100. That is, when the energy consumption N is the lowest, the second energy consumption score is recorded as 100 points. When the energy consumption N is the highest, the second energy consumption score is recorded as 0 points. The rest are calculated according to the formula. The second energy consumption score is between 0 and 100.

[0148] Where N is the energy consumption per unit of steel coil production, N=∑(Ne) / n, Nmin is the minimum energy consumption per unit of steel coil production, and Nmax is the minimum energy consumption per unit of steel coil production.

[0149] Step 4.25: Optimize the production plan model based on the cost per ton of steel produced from steel coils. The evaluation formula is: C' = ((C-Cmin) / (Cmax-Cmin))*100. That is, when the cost per ton of steel C is the lowest, the second cost score is 100 points; when the cost per ton of steel C is the highest, the second cost score is 0 points. The rest are calculated according to the formula, and the second cost score is between 0 and 100.

[0150] Where C is the cost per ton of steel produced from a steel coil, C=∑(Ce) / n, Cmin is the minimum cost per ton of steel produced from a steel coil, and Cmax is the minimum cost per ton of steel produced from a steel coil.

[0151] Step 4.26: Calculate the green design index of the new product based on weights. The calculation formula is as follows:

[0152] Green design index of new products = λ 61 *Second performance score +λ 62 *Second yield rate score +λ 63 *Second energy efficiency rating +λ 64 *Second cost rating.

[0153] Where, λ 61 λ is the weight for the second performance score. 62 λ is the weight for the second yield rate score. 63 λ is the weight of the second energy consumption score. 64 λ is the weight of the second cost score, and λ 61 +λ 62 +λ 63 +λ 64 =1.

[0154] Step 4.27: Calculate the deviation coefficient P between the chemical composition and process parameters in the new product production plan and the actual chemical composition and production process parameters of the most similar steel grade (i.e., the steel grade corresponding to the steel tapping mark with the smallest sum of normalized Euclidean distance DIST). The calculation formula is: P = P1 + P2.

[0155] Wherein, P1 is the first deviation coefficient between the chemical composition of the new product and the chemical composition of the most similar steel grade in the production plan, and the calculation formula is:

[0156]

[0157] Where, λ i Assigning the weights to each chemical component, x represents the design values ​​for each chemical component in the production plan of the new product. ik For the actual values ​​of each chemical composition of the most similar steel grade, N i denoted as the number of records for the most similar steel grade, and i represents the number of records for the chemical composition.

[0158] P2 is the second deviation coefficient between the process parameters in the new product production plan and the process parameters of the most similar steel grade. The calculation formula is:

[0159]

[0160] Where, β i Assigning weights to each process parameter. g represents the design values ​​of various process parameters in the production plan for the new product. ik N represents the actual values ​​of each process parameter for the most similar steel grade. i is the number of records for the most similar steel grade, and i is the number of process parameters.

[0161] Preferably, the chemical composition may include the content range of elements such as C, Si, Mn, Al, and P in the silicon steel product.

[0162] Step 4.28: Correct the green design index of the new product using the deviation coefficient P to obtain the corrected green design index of the new product. The correction calculation formula is as follows:

[0163] The corrected green design index of a new product = the green design index of the new product - P.

[0164] Example 1:

[0165] Step 1: Collect data and integrate user data, laboratory R&D data, and large-scale production data.

[0166] The user data includes user performance requirements for various types of silicon steel products (including electromagnetic properties such as iron loss and magnetic induction), user ID, inquiry ID, and basic user information.

[0167] The aforementioned laboratory R&D data includes chemical composition, R&D process parameters, various performance indicators (such as iron loss, magnetic induction, and other electromagnetic properties), and microstructure analysis results throughout the entire experimental process for various types of silicon steel products. The entire experimental process includes laboratory steelmaking, laboratory hot rolling, laboratory normalizing, laboratory cold rolling, laboratory continuous annealing, laboratory decarburization annealing, laboratory magnesium oxide coating, laboratory high-temperature annealing, and laboratory coating processes. The laboratory R&D data can be obtained from the silicon steel product R&D management system, and all data in the laboratory R&D data has been accumulated for at least one year to ensure the accuracy of the production plan.

[0168] The aforementioned large-scale production data comprises production data from each unit throughout the entire large-scale production process of various types of silicon steel products. This includes chemical composition (e.g., elements C, Si, S, etc.), tapping marks, production material tracking information (e.g., the corresponding unit number, coil entry number, and coil exit number during silicon steel coil production), actual production process parameters (e.g., tapping temperature, maximum normalizing temperature, etc., including low-frequency data collected by coil and high-frequency data collected by distance or time interval; each coil has only one low-frequency data point and multiple high-frequency data points), various performance data (e.g., iron loss, magnetic induction, and other electromagnetic properties), inspection and testing data, surface defect data, quality judgment data, coil weight at entry / exit, yield rate, coil weight allocation at entry, energy consumption per ton of steel, and cost per ton of steel. This large-scale production data can be obtained from the process control system and manufacturing execution system of steel production. All data points in this large-scale production data have been accumulated for at least one year to ensure the accuracy of the constructed solution model.

[0169] Step 2: Construct user data themes, laboratory R&D data themes, and large-scale production data themes.

[0170] Step 2.1: Construct a user demand data topic. The specific operation is as follows: Based on the user ID and inquiry ID in the user data, link the basic information of a user with the performance requirements of the user for silicon steel products to form a user demand data topic.

[0171] Step 2.2: Constructing a laboratory R&D data theme for silicon steel products, the specific steps of which are:

[0172] Step 2.2.1: Configure laboratory R&D data anomaly rules for the chemical composition, R&D process parameters, and performance indicators of silicon steel products.

[0173] Step 2.2.2: Based on the rules for abnormal laboratory R&D data, preprocess the laboratory R&D data. If the chemical composition, R&D process parameters, or performance indicators exceed their reasonable data range, the laboratory R&D data is judged to be abnormal, and the R&D personnel are notified that there is abnormal data, so that the R&D personnel can handle the abnormal data accordingly. Otherwise, the laboratory R&D data is judged to be normal.

[0174] Step 2.2.3: Based on the entire experimental process, according to the steelmaking date and heat number, and the slab and hot-rolled / cold-rolled coil information of silicon steel products, define the entry sample number and exit sample number for each experimental process. Using the logic that the exit sample number of the previous experimental process equals the entry sample number of the next experimental process, string together the entire experimental process data of silicon steel products to form a laboratory R&D data theme, which facilitates the traceability of the entire experimental process from steelmaking to cold rolling.

[0175] Step 2.3: Construct a large-scale production data theme for silicon steel products. The specific steps are as follows:

[0176] Step 2.3.1: Configure large-scale production data anomaly rules for the chemical composition, actual production process parameters, and performance data of silicon steel products.

[0177] Step 2.3.2: Based on the rules for anomalies in large-scale production data, preprocess the large-scale production data. If the actual chemical composition, production process parameters, and performance data exceed their reasonable range, they are judged as abnormal large-scale production data, and the R&D personnel are notified that there is abnormal data, so that the R&D personnel can handle the abnormal data accordingly. Otherwise, the large-scale production data is judged as normal.

[0178] Step 2.3.4: Calculate the average, maximum, minimum and other characteristic values ​​of the high-frequency data and performance data of the silicon steel coil.

[0179] Step 2.3.5: Based on the entire large-scale production process, define the exit roll number and inlet roll number of each unit. Using the logic that the exit roll number of the previous unit equals the inlet roll number of the next unit, connect the large-scale production process data of silicon steel products to form a large-scale production data theme.

[0180] The large-scale production process data includes actual production process parameters, inspection and testing data, surface defect data, quality judgment data, yield, energy consumption per ton of steel, and cost per ton of steel for each unit. This allows for the traceability of the entire large-scale production process of any steel coil, starting from any process, and includes all data such as actual production process parameters and inspection and testing results for each unit.

[0181] Step 3: Input the required information such as the type, specifications and performance of the silicon steel product. In this example, the user's required information is: oriented silicon steel, with a thickness of 0.20mm, and an electromagnetic induction strength B8 (magnetic polarization strength corresponding to a magnetic field strength of 800A / m) of not less than 1.93T.

[0182] Based on user needs, the data for large-scale production and R&D is filtered separately, with an adjustment factor of 1.01, allowing for the acquisition of the corresponding data.

[0183] The matched data were standardized, and the similarity between the matched data and the magnetic performance requirements was calculated. The similarity was 1.93 for large-scale production data and 1.95 for R&D data. The electromagnetic performance data used the mean of each sampling point.

[0184] Set the performance similarity threshold to 50%, filter the data from the two categories that meet the performance similarity, and mark the products that meet the similarity threshold as optimized products. Then generate the production plan for the optimized products.

[0185] The R&D team selected the elemental Si content as the key chemical component and the decarburization temperature range as the key process parameter, with control precisions of 0.05 and 5, respectively. Based on these control precisions, five classification schemes can be obtained, as shown in Table 1.

[0186] Table 1. Classification scheme based on chemical composition (Si content) and decarburization temperature range

[0187] Si content range Decarbonization temperature range Scheme Number (3.20,3.25] (840,845] 1 (3.20,3.25] (835,840] 2 (3.20,3.25] (830,835] 3 [3.15,3.25] (835,840] 4 [3.15,3.25] (830,835] 5

[0188] Step 4: Conduct a multi-dimensional evaluation of the five proposed solutions to obtain green manufacturing indicators and recommend a solution:

[0189] Performance rating: The performance of production plan 1 matches the performance data of the corresponding large production data, and the first performance score is 100 points. The rest are scored based on similarity.

[0190] Yield Scoring: The yield of this product is assessed based on the normalizing, rolling, and decarburizing mills, with decarburization having a higher weighting. First Yield = 0.3 * Normalizing Mill Yield + 0.3 * Rolling Mill Yield + 0.4 * Decarburizing Mill Yield.

[0191] First energy consumption score: A comprehensive calculation is performed based on the energy consumption per ton of steel selected from the data topics.

[0192] First cost score: A comprehensive calculation is performed based on the cost per ton of steel for the selected data topics.

[0193] Green Design Index: Green Design Index = 0.5 * First Performance Score + 0.3 * First Material Yield Score + 0.1 * First Energy Consumption Score + 0.1 * First Cost Score. The detailed score results are shown in Table 2.

[0194] Table 2. Multidimensional evaluation of the five schemes and their green design indices

[0195]

[0196] As shown in Table 2, Option 1 has the highest score, so Option 1 is recommended as the preferred option.

[0197] Example 2:

[0198] Step 1: Collect data and integrate user data, laboratory R&D data, and large-scale production data.

[0199] Step 2: Construct user data themes, laboratory R&D data themes, and large-scale production data themes.

[0200] Step 3: Input the required information such as the type, specifications, and performance of the silicon steel product. In this example, the user's requirements are: grain-oriented silicon steel, thickness 0.20mm, electromagnetic induction intensity B8 (magnetic polarization intensity corresponding to a magnetic field strength of 800A / m) not less than 1.93T, and iron loss P. 1.7 / 50 Products not exceeding 1.0 W / kg.

[0201] Based on user demand information, the data themes of large-scale production and R&D are filtered separately. The adjustment factors for magnetic induction and iron loss are 1.01 and 0.98, respectively, so that the corresponding data can be obtained.

[0202] The matched data were standardized, and the similarity between the matched data and the magnetic performance requirements was calculated. The corresponding values ​​for large-scale production data were 1.93 for magnetic induction and 1.0 for iron loss, while the corresponding values ​​for R&D data were 1.95 for magnetic induction and 0.98 for iron loss. The magnetic performance data were calculated using the mean of each sampling point.

[0203] Set the performance similarity threshold to 50%, filter the data that meets the performance similarity between the two types of data. If there are no products that meet the similarity threshold, mark them as new products. Then generate the production plan for the new products.

[0204] We selected data on grain-oriented silicon steel products with a thickness of 0.02 mm from the past year's production data.

[0205] The product developers selected the key chemical components as element C and Si content, and the key process parameter as decarburization temperature, with control precisions of 0.002, 0.05, and 5, respectively. Based on the data from the previous steps, xgboost (maximum tree depth of 5), neural network (two hidden layers), training inputs (C content, Si content, decarburization temperature), and outputs (magnetic induction, iron loss) were used for evaluation. The model trained using xgboost (maximum tree depth of 5) showed the smallest mean absolute error and was thus the optimal performance indicator prediction model.

[0206] The baselines for obtaining C content, Si content, and decarburization temperature based on large-scale production data are (0.056, 3.25, 845).

[0207] Five schemes were formed by gridded search based on the benchmark points, as shown in Table 3.

[0208] Table 3. Classification scheme based on chemical composition (C and Si content) and decarburization temperature range

[0209] C content range Si content range Decarbonization temperature range Scheme Number (0.056,0.058] (3.25,330] (845,850] 1 (0.056,0.058] (3.25,330] [840,845] 2 (0.056,0.058] [3.20,3.25] (845,850] 3 [0.054,0.056] [3.20,3.25] (845,850] 4 [0.054,0.056] [3.20,3.25] [840,845] 5

[0210] Predict magnetic induction and iron loss performance based on the optimal performance index prediction model.

[0211] Step 4: Multi-dimensional Solution Evaluation: A multi-dimensional evaluation of the five solutions developed above is conducted to obtain green manufacturing indicators, and a recommended solution is then provided.

[0212] Performance rating: None of the above 5 schemes have been produced, meaning they cannot match the actual performance of mass production. The calculation formula is: Second performance rating = 0.4 * magnetic similarity + 0.6 * iron loss similarity;

[0213] Second yield rate score: The yield rate of this product focuses on the normalizing, decarburizing, and continuous annealing units, among which decarburizing has a higher weight. Second yield rate = 0.3 * normalizing unit yield rate + 0.4 * decarburizing unit yield rate + 0.3 * continuous annealing unit yield rate.

[0214] Second energy consumption score: A comprehensive calculation is performed based on the energy consumption per ton of steel selected from the data topics.

[0215] Second cost score: A comprehensive calculation is performed based on the cost per ton of steel for the selected data topics.

[0216] Green Design Index: Green Design Index = 0.5 * Second Performance Score + 0.3 * Second Material Yield Score + 0.1 * Second Energy Consumption Score + 0.1 * Second Cost Score.

[0217] Calculation of deviation coefficient:

[0218]

[0219] Table 4. Multidimensional evaluation of the five schemes and their green design indices

[0220]

[0221] As shown in Table 4, Option 4 has the highest score, so Option 4 is recommended as the preferred option.

[0222] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A smart design evaluation method for silicon steel product manufacturing schemes, characterized by: Includes the following steps: Step 1: Collect data and integrate user data, laboratory R&D data, and large-scale production data; The user data includes users' performance requirements for various types of silicon steel products, user ID, inquiry ID, and basic user information; The laboratory research and development data mentioned above includes chemical composition, research and development process parameters, various performance indicators, and microstructure analysis results throughout the entire experimental process of various types of silicon steel products; The aforementioned large-scale production data refers to the production data of each unit in the entire large-scale production process of various types of silicon steel products, including chemical composition, steel tapping marks, production material tracking information, actual production process parameters, various performance data, inspection and testing data, surface defect data, quality judgment data, inlet / outlet coil weight, yield rate, inlet coil weight allocation, energy consumption per ton of steel, and cost per ton of steel. Step 2: Construct user data themes, laboratory R&D data themes, and large-scale production data themes; Step 2.1: Construct the user demand data theme; Step 2.2: Construct laboratory R&D data themes for silicon steel products; Step 2.3: Construct a large-scale production data theme for silicon steel products; Step 3: Input the requirements for the type, specifications, and performance of silicon steel products, and generate several production plans for silicon steel products based on the requirements. The production plans include production plans for optimized products and production plans for new products. Step 4: Conduct a multi-dimensional comprehensive evaluation of the production plan for each silicon steel product generated in Step 3 to obtain the green design index of the silicon steel product production plan, and recommend a production plan based on the green design index; In step 4, the evaluation dimensions include performance similarity, yield, energy consumption per ton of steel, and cost per ton of steel. For each of the aforementioned optimized product production plans, the multi-dimensional comprehensive evaluation method is as follows: Step 4.11: Calculate the second performance similarity between the performance data of the optimized product's production plan and the actual performance data of mass production. If the second performance similarity is fully satisfied, that is, the performance data of the optimized product's production plan and the actual performance data of mass production are completely matched, then the first performance score is recorded as 100 points, and proceed to step 4.

13. If the second performance similarity is not fully satisfied, then proceed to step 4.

12. Step 4.12: Calculate the similarity between the magnetic induction and iron loss of the optimized product's production plan and the actual magnetic induction and iron loss in mass production, and perform performance evaluation based on weights to obtain the third performance similarity S3. The evaluation formula is: S3 = λ 11 *S 磁感1 +λ 21 *S 铁损1 The first performance score is recorded as S3; Among them, S 磁感1 To optimize the similarity between the magnetic properties of the product's manufacturing performance data and the actual magnetic properties of its mass production performance data, λ 11 For S 磁感1 The weights of S; 铁损1 To optimize the similarity between the iron loss data of the product's production plan performance data and the iron loss data of actual mass production performance data, λ 21 For S 铁损1 The weights; λ 11 +λ 21 =1; Step 4.13: Based on the weighted evaluation of the production scheme model of the product, the first yield rate of each unit is evaluated and optimized to obtain the first yield rate index; The formula for calculating the first yield of each unit is: First yield rate η1 = (export weight / weight of imported coil) * 100; Wherein, export weight = ∑(weight of each steel coil exported), and the weight of each steel coil allocated to the inlet coil = ∑(weight of each steel coil allocated to the inlet coil); The formula for calculating the first yield rate index is: First yield rate index = ∑(λ3*η1), where λ3 is the weight of the first yield rate corresponding to each unit, and ∑λ3 = 1. The score of the first yield rate is recorded as the first yield rate index. Step 4.14: Optimize the energy consumption of the production scheme model based on the energy consumption of steel coil production unit. When the energy consumption is the lowest, the first energy consumption score is recorded as 100 points; when the energy consumption is the highest, the first energy consumption score is recorded as 0 points. Step 4.15: Based on the cost per ton of steel produced from steel coils, evaluate and optimize the production plan model for the product. When the cost per ton of steel is the lowest, the first cost score is recorded as 100 points; when the cost per ton of steel is the highest, the first cost score is recorded as 0 points. Step 4.16: Calculate the green design index of the optimized product based on weights. The calculation formula is as follows: Optimize the product's green design index = λ 41 *First performance score +λ 42 *First yield score +λ 43 *First Energy Efficiency Rating +λ 44 *First cost rating; Where, λ 41 λ is the weight for the first performance score. 42 λ is the weight for the first yield rate score. 43 λ is the weight of the first energy consumption score. 44 Let λ be the weight of the first cost score, and λ be the weight of the first cost score. 41 +λ 42 +λ 43 +λ 44 =1.

2. The intelligent design evaluation method for silicon steel product manufacturing schemes according to claim 1, characterized in that: The production process parameters recorded include low-frequency data collected by steel coil and high-frequency data collected by distance or time interval. Each steel coil has one low-frequency data point and multiple high-frequency data points.

3. The intelligent design evaluation method for silicon steel product manufacturing schemes according to claim 1, characterized in that: Step 2.1 includes: based on the user ID and inquiry ID in the user data, linking the basic information of a user with the performance requirements of the user for silicon steel products to form a user demand data theme; Step 2.2 includes: Step 2.2.1: Configure laboratory R&D data anomaly rules for the chemical composition, R&D process parameters, and performance indicators of silicon steel products; Step 2.2.2: Preprocess the laboratory R&D data based on the anomaly rules of the laboratory R&D data; Step 2.2.3: Based on the entire experimental process, define the entry sample number and exit sample number for each experimental procedure. Using the logic that the exit sample number of the previous experimental procedure equals the entry sample number of the next experimental procedure, connect the entire experimental process data of silicon steel products to form the laboratory R&D data theme. Step 2.3 includes: Step 2.3.1: Configure large-scale production data anomaly rules for the chemical composition, actual production process parameters, and performance data of silicon steel products; Step 2.3.2: Preprocess the large-scale production data based on the anomaly rules of the large-scale production data; Step 2.3.4: Calculate the high-frequency data and performance characteristic values ​​of the silicon steel coils; Step 2.3.5: Based on the entire large-scale production process, define the exit roll number and inlet roll number of each unit. Using the logic that the exit roll number of the previous unit equals the inlet roll number of the next unit, connect the large-scale production process data of silicon steel products to form a large-scale production data theme.

4. The intelligent design evaluation method for silicon steel product manufacturing schemes according to claim 1, characterized in that: Step 3 includes: Step 3.1: Filter the laboratory R&D data theme and the large-scale production data theme according to the demand information, and select the two types of relevant data that meet the demand information; among them, the performance indicators in the laboratory R&D data theme are multiplied by the adjustment factor before filtering; Step 3.2: Based on the performance requirements in the demand information, normalize and standardize the data for each type of relevant data; Step 3.3: For each type of relevant data, calculate the first performance similarity between the performance requirements and the normalized and standardized performance data; Step 3.4: Set a performance similarity threshold. If the first performance similarity is greater than or equal to the performance similarity threshold, then mark the silicon steel product corresponding to the first performance similarity as an optimized product and proceed to step 3.

5. If the first performance similarity is less than the performance similarity threshold, then mark the silicon steel product corresponding to the first performance similarity as a new product and proceed to step 3.

6. Step 3.5: Generate an optimized production plan for the product: Combine the actual production process parameters corresponding to the first performance similarity with the control precision of each unit in mass production, and divide the actual production process parameters. Step 3.6: Based on performance requirements, establish a performance index prediction model and verify the optimal production plan model for the new product.

5. The intelligent design evaluation method for silicon steel product manufacturing schemes according to claim 4, characterized in that: Step 3.6 includes: Step 3.6.1: Select historical data from the large production data. The types and specifications of silicon steel products in this historical data are the same as those of the input silicon steel products, and the time range of the large production data is the most recent year. Step 3.6.2: Establish several performance index prediction models; the input of each performance index prediction model is the key chemical components and key process parameters selected by the R&D personnel, and the output is the performance index. The optimal performance prediction model among several performance index prediction models is determined through cross-validation. Step 3.6.3: Calculate the average value of the actual performance of the chemical composition and production process parameters in the large-scale production data selected in Step 3.6.1, and use the average value of the actual performance of the chemical composition and production process parameters as the benchmark point of the scheme; Step 3.6.4: Perform a gridded search on the performance index prediction model starting from the baseline of the scheme.

6. The intelligent design evaluation method for silicon steel product manufacturing schemes according to claim 1, characterized in that: For each of the aforementioned new product production plans, the multi-dimensional comprehensive evaluation method is as follows: Step 4.21: Calculate the fourth performance similarity between the performance data of the production plan of the new product and the actual performance data of mass production. If it is completely satisfied, that is, the performance data of the production plan of the new product and the actual performance data of mass production are completely matched, then the second performance score is recorded as 100 points, and proceed to step 4.

23. If it is not completely satisfied, then execute step 4.

22. Step 4.22: Calculate the similarity between the magnetic induction and iron loss of the production plan for the new product and the actual magnetic induction and iron loss in mass production, and perform performance evaluation based on weights to obtain the fifth performance similarity S5. The evaluation formula is: S5 = λ 12 *S 磁感2 +λ 22 *S 铁损2 The second performance rating is recorded as S5. Among them, S 磁感2 λ represents the similarity between the magnetic properties of the production plan performance data and the actual magnetic properties of the mass production performance data for a new product. 12 For S 磁感2 The weights of S; 铁损2 λ represents the similarity between the iron loss data of the production plan and the iron loss data of the actual mass production. 22 For S 铁损2 The weights; λ 12 +λ 22 =1; Step 4.23: Evaluate the second yield rate of each unit in the production plan of the new product based on the weight, and obtain the second yield rate index; The formula for calculating the second yield rate of each unit is: Second yield rate η2 = (export weight / weight of imported coil) * 100; Wherein, export weight = ∑(weight of each steel coil exported), and the weight of each steel coil allocated to the inlet coil = ∑(weight of each steel coil allocated to the inlet coil); The formula for calculating the second yield rate index is: Second yield rate index = ∑(λ5*η2), and ∑λ5 = 1. The score of the second yield rate is recorded as the second yield rate index. Step 4.24: Evaluate the energy consumption of the production plan for the new product based on the unit energy consumption of steel coil production. When the energy consumption is the lowest, the second energy consumption score is recorded as 100 points; when the energy consumption is the highest, the second energy consumption score is recorded as 0 points. Step 4.25: Based on the cost per ton of steel produced from steel coils, evaluate and optimize the production plan for the product. When the cost per ton of steel is the lowest, the second cost score is recorded as 100 points; when the cost per ton of steel is the highest, the second cost score is recorded as 0 points. Step 4.26: Calculate the green design index of the new product based on weights. The calculation formula is as follows: Green design index of new products = λ 61 *Second performance score +λ 62 *Second yield rate score +λ 63 *Second energy efficiency rating +λ 64 *Second cost rating; Where, λ 61 λ is the weight for the second performance score. 62 λ is the weight for the second yield rate score. 63 λ is the weight of the second energy consumption score. 64 λ is the weight of the second cost score. 61 +λ 62 +λ 63 +λ 64 =1; Step 4.27: Calculate the deviation coefficient P = P1 + P2 between the new product production plan and the most similar steel grade; Wherein, P1 is the first deviation coefficient between the chemical composition of the new product production plan and the chemical composition of the most similar steel grade, and the calculation formula is: Where, λ i Assigning the weights to each chemical component, x represents the design values ​​for each chemical component in the production plan of the new product. ik For the actual values ​​of each chemical composition of the most similar steel grade, N i The number of records for the most similar steel grade, where i is the number of chemical compositions; P2 is the second deviation coefficient between the process parameters in the new product production plan and the process parameters of the most similar steel grade. The calculation formula is as follows: Where, β i Assigning weights to each process parameter. g represents the design values ​​of various process parameters in the production plan for the new product. ik N represents the actual values ​​of each process parameter for the most similar steel grade. i The number of records for the most similar steel grade, where i is the number of process parameters; Step 4.28: Correct the green design index of the new product using the deviation coefficient P to obtain the corrected green design index of the new product. The correction calculation formula is as follows: The corrected green design index of a new product = the green design index of the new product - P.

Images