Cold-rolled product quality component multi-objective optimization method based on data analysis

Abstract

The invention provides a cold-rolled product quality component multi-objective optimization method based on data analysis, and belongs to the field of material design, and the method comprises the steps: inputting a plurality of performance indexes of the quality of a to-be-analyzed cold-rolled product into a quality analysis and multi-objective optimization model, and obtaining the component data of the cold-rolled product corresponding to the plurality of indexes, an establishment process of the quality analysis and multi-objective optimization model comprises the following steps: based on historical data of a cold rolling production process, constructing a corresponding relationship between a plurality of indexes of the quality of a cold rolling product and component data of the cold rolling product by adopting a full-connection neural network, and adopting a fractal variation strategy to introduce a multi-objective differential evolution algorithm to carry out reverse optimization on the corresponding relationship, and obtaining a quality analysis and multi-objective optimization model. The method has good practicability and stability, and the quality components of the cold-rolled product can be effectively optimized.

Description

Technical Field

[0001] This invention belongs to the field of materials design, specifically relating to a multi-objective optimization method for the quality composition of cold-rolled products based on data analysis. Background Technology

[0002] For cold-rolled products, their surface quality and mechanical properties are macroscopic representations of their internal microstructure, and the formation process of different microstructures is constrained by chemical composition. The performance of cold-rolled products is influenced by chemical elements such as carbon. Based on carbon content, cold-rolled products can be classified as high-carbon, medium-carbon, low-carbon, and ultra-low-carbon. Furthermore, to meet the requirements of steel performance, structure, and application conditions, various alloying elements need to be added to cold-rolled products. Manganese is an excellent deoxidizer and desulfurizer, improving the plasticity, machinability, and deformability of cold-rolled products during heat treatment, processing, and forming. Chromium and silicon can significantly improve the corrosion resistance, heat resistance, and oxidation resistance of cold-rolled products, respectively. The addition of small amounts of aluminum can refine grains and improve impact toughness, oxidation resistance, and corrosion resistance. Rare earth elements can improve mechanical properties, corrosion resistance, wear resistance, and processing performance. Therefore, rationally designing the chemical composition of cold-rolled products is fundamental to improving their quality and performance.

[0003] Traditional experimental trial-and-error methods based on theoretical knowledge are time-consuming and costly for cold-rolled product composition design, and can no longer meet practical needs. In recent years, with the continuous improvement of cold-rolling production process data, data-driven artificial intelligence methods can comprehensively consider the nonlinear relationships between different cold-rolled product characteristics and establish the correspondence between chemical composition and various properties, showing significant advantages in cold-rolled product performance prediction. In addition, high-performance optimization algorithms for single and multi-objective problems are constantly being proposed, and embedding data-driven composition quality analytical models as objective functions into optimization algorithms for composition design is gradually becoming a research hotspot in the field of quality composition design.

[0004] Currently, research on the quality of cold-rolled products is relatively scarce, and existing technologies still have shortcomings in terms of accuracy and stability in multi-objective optimization and product performance prediction. Chinese patent CN116542153A proposes a reinforcement learning-based inverse design method. Although this method is mainly used in metal processing, it suffers from stability issues in the design of cold-rolled material composition, potentially leading to insufficient training or model collapse. Furthermore, this method is highly dependent on data; insufficient or poor-quality data can severely affect the accuracy of the results. Chinese patent CN116796627A uses deep learning technology to achieve high-precision real-time prediction of cold-rolled steel sheet performance, comprehensively considering multiple factors and exhibiting high reliability and practicality based on data-driven approaches. However, this method also suffers from strong data dependence, limited model generalization ability, requires extensive experimental verification, and demands significant computational resources. Existing multi-objective optimization algorithms also struggle to effectively balance multiple performance objectives of cold-rolled products. Therefore, a new method is urgently needed that can improve both efficiency and quality while optimizing the quality composition of cold-rolled products. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application proposes a multi-objective optimization method for the quality composition of cold-rolled products based on data analysis.

[0006] Firstly, this application proposes a multi-objective optimization method for the quality composition of cold-rolled products based on data analysis, including:

[0007] Obtain multiple performance indicators of the cold-rolled product quality to be analyzed;

[0008] Multiple performance indicators of the cold-rolled product quality to be analyzed are input into the quality analysis and multi-objective optimization model to obtain the composition data of the cold-rolled product corresponding to the multiple indicators. The quality analysis and multi-objective optimization model is established by: constructing the correspondence between the multiple indicators of the cold-rolled product quality and the composition data of the cold-rolled product based on historical data of the cold-rolling production process using a fully connected neural network; and using a fractal mutation strategy to introduce a multi-objective differential evolution algorithm to perform reverse optimization on the correspondence to obtain the quality analysis and multi-objective optimization model.

[0009] The method, based on historical data from the cold rolling production process, employs a fully connected neural network to construct the correspondence between multiple indicators of cold-rolled product quality and the composition data of cold-rolled products, including:

[0010] Obtain historical production data for each batch during the cold rolling process and preprocess the historical production data;

[0011] Feature extraction was performed on the preprocessed historical production data, and the extracted features were correlated with multiple performance indicators of cold-rolled products.

[0012] A fully connected neural network was used to construct the correspondence between multiple indicators of cold-rolled product quality and the extracted features.

[0013] The aforementioned performance indicators include: yield strength, tensile strength, and elongation of cold-rolled products.

[0014] The step of employing a fractal mutation strategy to introduce a multi-objective differential evolution algorithm to perform reverse optimization of the correspondence includes:

[0015] Step S2.4: Select N different combinations of performance indicators for yield strength, tensile strength and elongation of cold-rolled products from the historical production data of the cold rolling production process, and record the composition data of the actual cold-rolled products corresponding to each combination of performance indicators.

[0016] Step S2.5: Input each set of performance indexes into a fully connected neural network to obtain the predicted composition data of the cold-rolled product;

[0017] Step S2.6: Using the predicted composition data of the cold-rolled product as individuals in the population, a multi-objective differential evolution algorithm is introduced using a fractal mutation strategy to optimize the individuals in the population, thereby obtaining the optimized predicted composition data of the cold-rolled product.

[0018] Step S2.7: Calculate the absolute deviation between the optimized predicted composition data of the cold-rolled product and the actual composition data of the cold-rolled product;

[0019] Step S2.8: If the absolute deviation is greater than the preset threshold, change the predicted composition data of the cold-rolled product and return to step S2.6 to re-optimize;

[0020] Step S2.9: If the absolute deviation is less than or equal to the preset threshold, stop the iteration and output the optimized predicted composition data of the cold-rolled product.

[0021] The process involves using the predicted composition data of cold-rolled products as individuals in a population, employing a fractal mutation strategy to introduce a multi-objective differential evolution algorithm to optimize the individuals in the population, resulting in optimized predicted composition data of the cold-rolled products, including:

[0022] Step S2.6.1: Initialize the individuals in the population by using the predicted composition data of the cold-rolled product as individuals in the population;

[0023] Step S2.6.2: Use a multi-objective differential evolution algorithm with fractal mutation strategy to perform differential mutation operation on the initialized individuals to obtain mutated individuals;

[0024] Step S2.6.3: Perform a crossover operation on the mutated individuals to generate offspring individuals after the crossover operation;

[0025] Step S2.6.4: If the parent individual dominates the offspring individual it produces, the parent individual will be retained. If the offspring individual dominates the parent individual, the offspring individual will be retained. If neither dominates the other, both will be retained. If the parent individual dominates the offspring individual, the parent individual will be retained and the offspring individual will be eliminated. If the offspring individual dominates the parent individual, the offspring individual will be retained. If neither the parent individual nor the offspring individual dominates the parent individual, both the parent individual and the offspring individual will be retained.

[0026] Step S2.6.5: If the current iteration count reaches the maximum iteration count, perform non-dominated sorting and use the first level after sorting as the optimized predicted composition data of the cold-rolled product; if the current iteration count does not reach the maximum iteration count, return to step S2.6.2 and continue the differential mutation operation.

[0027] The multi-objective differential evolution algorithm employing a fractal mutation strategy performs differential mutation operations on the initialized individuals, including:

[0028] Step S2.6.2.1: Randomly generate the maximum number of iterations from the interval [preset lower limit of maximum number of iterations, preset upper limit of maximum number of iterations];

[0029] Step S2.6.2.2: Take the initialized individual as the individual to be mutated, and use the iterative function system to generate the final coordinates of the individual to be mutated starting from the origin. ;

[0030] Step S2.6.2.3: Select three distinct individuals from the individuals to be mutated, i.e., the first individual. The second body And the third body ;

[0031] Step S2.6.2.4: Select two components from the final coordinates with equal probability, and combine them with the first volume... The second body And the third body , generating a mutated individual v.

[0032] The equal probability selection of two components in the final coordinates, combined with the first volume The second body And the third body Generate a mutated individual v, calculated as follows:

[0033] ;

[0034] Where v represents the mutated individual, x represents the final x-value of the individual to be mutated, and y represents the final x-value of the individual to be mutated. For the first body, For the second body, For the third entity.

[0035] Secondly, this application proposes an electronic device, comprising: one or more processors, and a memory for storing instructions, which, when executed by the one or more processors, cause the one or more processors to perform the aforementioned multi-objective optimization method for the quality composition of cold-rolled products based on data parsing.

[0036] Thirdly, this application proposes a computer-readable storage medium storing executable instructions that, when executed, cause a processor to perform the aforementioned multi-objective optimization method for the quality composition of cold-rolled products based on data analysis.

[0037] Fourthly, this application proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned multi-objective optimization method for the quality composition of cold-rolled products based on data analysis.

[0038] Beneficial effects:

[0039] Based on actual production data from a steel plant, this application proposes a multi-objective optimization method for the quality composition of cold-rolled products using data analysis. This method analyzes the chemical composition of cold-rolled products using a Fully Connected Neural Network (FCNN) model, treating the composition as a decision variable and the difference between predicted and target performance as the objective function. An improved Multiobjective Differential Evolution Algorithm (MODE) is then used for composition optimization. Compared with actual production data, this method demonstrates good practicality and stability, effectively optimizing the quality composition of cold-rolled products. Attached Figure Description

[0040] Figure 1 A flowchart of a multi-objective optimization method for the quality composition of cold-rolled products based on data analysis, according to an embodiment of this application;

[0041] Figure 2 A schematic diagram of the optimized predicted composition data flow of cold-rolled products according to an embodiment of this application;

[0042] Figure 3 A schematic diagram of the differential mutation process in an embodiment of this application;

[0043] Figure 4 Example diagram of fractal patterns in embodiments of this application. Detailed Implementation

[0044] The specific implementation methods of this application will be further described in detail below with reference to the accompanying drawings and embodiments.

[0045] The surface quality and mechanical properties of cold-rolled products are macroscopic representations of their internal microstructure, and the formation process of different microstructures is constrained by chemical composition. Traditional methods for designing the composition of cold-rolled products rely on theoretical knowledge and experimental trial-and-error, which are time-consuming and costly, failing to meet actual production needs. With the continuous improvement of cold-rolling production process data, data-driven artificial intelligence methods can comprehensively consider the nonlinear relationships between different characteristics and establish the correspondence between chemical composition and various properties, showing significant advantages in performance prediction. Therefore, this application proposes a multi-objective optimization method for the quality composition of cold-rolled products based on data analysis. First, an FCNN is used to establish a predictive model between the chemical composition and quality performance of cold-rolled products. Then, a fractal mutation strategy is introduced into a multi-objective differential evolution algorithm to improve optimization performance. Finally, the improved optimization algorithm is used for reverse design to optimize the composition of cold-rolled products to meet the target performance.

[0046] Example 1:

[0047] This embodiment proposes a multi-objective optimization method for the quality composition of cold-rolled products based on data analysis, such as... Figure 1 As shown, it includes:

[0048] Step S1: Obtain multiple performance indicators of the quality of the cold-rolled product to be analyzed;

[0049] Step S2: Input multiple performance indicators of the cold-rolled product quality to be analyzed into the quality analysis and multi-objective optimization model to obtain the composition data of the cold-rolled product corresponding to the multiple indicators. The quality analysis and multi-objective optimization model is established by: using a fully connected neural network to construct the correspondence between the multiple indicators of the cold-rolled product quality and the composition data of the cold-rolled product based on historical data of the cold-rolled production process; and using a fractal mutation strategy to introduce a multi-objective differential evolution algorithm to perform reverse optimization on the correspondence to obtain the quality analysis and multi-objective optimization model.

[0050] The method, based on historical data from the cold rolling production process, employs a fully connected neural network to construct the correspondence between multiple indicators of cold-rolled product quality and the composition data of cold-rolled products, including:

[0051] Step S2.1: Obtain historical production data for each batch during the cold rolling process and preprocess the historical production data;

[0052] In this embodiment, firstly, historical production data for each batch during the cold rolling process is collected. The dataset size is 50,000 records, and each data point includes the content of 18 chemical elements and three performance indicators. The three performance indicators are: yield strength, tensile strength, and elongation of the cold-rolled product. Then, all historical production data are standardized by converting them to zero mean and unit standard deviation.

[0053] Step S2.2: Feature extraction is performed on the preprocessed historical production data. The extracted features are correlated with multiple performance indicators of cold-rolled product quality, including:

[0054] Step S2.2.1: Construct a feature matrix X, which contains the content of 18 chemical elements in all samples (18-dimensional features for each sample).

[0055] Step S2.2.2: Construct a label matrix Y, which contains the three performance indicators (3-dimensional label for each sample) of all samples: yield strength (YS), tensile strength (TS) and elongation (EL), as shown in Table (1);

[0056] Table 1. Upper and lower limits of the content of different chemical elements

[0057] Chemical elements The Lower World Upper Realm Chemical elements The Lower World Upper Realm Chemical elements The Lower World Upper Realm C 0.0011 0.06 Ni 0.01 0.03 Alt 0.021 0.061 Mn 0.1 0.41 Cu 0.02 0.06 Nb 0.001 0.003 S 0.001 0.031 Mo 0.003 0.014 N 0.0008 0.0089 P 0.003 0.032 V 0 0.004 As 0.0039 0.0155 Si 0 0.04 Ti 0.001 0.114 Ca 0 0.0031 Cr 0.01 0.06 Als 0.019 0.059 B 0 0.001

[0058] Step S2.2.3: Perform correlation analysis on each feature in the feature matrix X and the three performance indicators to show the correlation between each feature and the performance indicators.

[0059] Step S2.3: Use a fully connected neural network to construct the correspondence between multiple indicators of cold-rolled product quality and the extracted features.

[0060] In this embodiment, an initial quality analysis and multi-objective optimization model is constructed using a fully connected neural network (FCNN). This model establishes a quality analysis model between chemical composition and quality performance. The FCNN consists of an input layer with 18 neurons, an output layer with 3 neurons, and four hidden layers with 32, 24, 16, and 8 neurons respectively. This model predicts the corresponding performance indicators based on the given chemical composition. The model is then embedded as an objective function into the optimization algorithm for composition design.

[0061] The method employs a fractal mutation strategy to introduce a multi-objective differential evolution algorithm to perform reverse optimization of the correspondence, as shown, including:

[0062] Step S2.4: Select N different combinations of performance indicators for yield strength, tensile strength and elongation of cold-rolled products from the historical production data of the cold rolling production process, and record the composition data of the actual cold-rolled products corresponding to each combination of performance indicators.

[0063] Step S2.5: Input each set of performance indexes into a fully connected neural network to obtain the predicted composition data of the cold-rolled product;

[0064] Step S2.6: Using the predicted composition data of the cold-rolled product as individuals in the population, a multi-objective differential evolution algorithm is introduced using a fractal mutation strategy to optimize the individuals in the population, thereby obtaining the optimized predicted composition data of the cold-rolled product.

[0065] Step S2.7: Calculate the absolute deviation between the optimized predicted composition data of the cold-rolled product and the actual composition data of the cold-rolled product;

[0066] Step S2.8: If the absolute deviation is greater than the preset threshold, change the predicted composition data of the cold-rolled product and return to step S2.6 to re-optimize;

[0067] Step S2.9: If the absolute deviation is less than or equal to the preset threshold, stop the iteration and output the optimized predicted composition data of the cold-rolled product.

[0068] The predicted composition data of the cold-rolled product is used as individuals in a population. A fractal mutation strategy is employed to introduce a multi-objective differential evolution algorithm to optimize the individuals in the population, resulting in optimized predicted composition data of the cold-rolled product. Figure 2 As shown, it includes:

[0069] Step S2.6.1: Initialize the individuals in the population by using the predicted composition data of the cold-rolled product as individuals in the population;

[0070] In this embodiment, an initial population for pruning using a multi-objective differential evolution algorithm is obtained based on the length of the fully connected neural network genotype data and the preset population size, and each individual is evaluated and its objective function value is calculated.

[0071] In this embodiment, the genotype data length refers to the dimension of the binary vector used to encode the structural parameters of the fully connected neural network. Its specific value is determined by the number of network layers, the number of neurons in each layer, and the connection method. For example, when selectively retaining / deleting neurons in a specific layer of the network, the genotype data length corresponds to the total number of parameters that can be pruned in that layer. Based on the genotype data length and the preset population size, an initial population for the multi-objective differential evolution algorithm is generated, and each individual is evaluated to calculate its objective function value.

[0072] Step S2.6.2: A multi-objective differential evolution algorithm using a fractal mutation strategy is employed to perform differential mutation operations on the initialized individuals, resulting in mutated individuals, such as... Figure 3 As shown, it includes:

[0073] Step S2.6.2.1: Randomly generate the maximum number of iterations from the interval [preset lower limit of maximum number of iterations, preset upper limit of maximum number of iterations];

[0074] In this embodiment, the maximum number of iterations is randomly generated from the interval. The maximum number of iterations is randomly generated, and the final coordinates are generated starting from the origin using the IFS of the individual to be mutated. .

[0075] Step S2.6.2.2: Take the initialized individual as the individual to be mutated, and use the iterative function system to generate the final coordinates of the individual to be mutated starting from the origin. ;

[0076] In this embodiment, the initialized individual is taken as the individual to be mutated, and the final coordinates of the individual to be mutated are generated simultaneously along the x-axis and y-axis in the xy two-dimensional plane starting from the origin (0,0) of the two-dimensional coordinate system using an iterative function system. The maximum number of iterations is randomly generated from the interval, and the final coordinates are generated starting from the origin using the IFS corresponding to the individual to be mutated. ,like Figure 4 As shown, the process of generating the graphic in step S2.6.2.2 is existing technology and will not be described again in this embodiment.

[0077] In this embodiment, the fractal mutation strategy is to use an Iterated Function System (IFS) to generate a two-dimensional fractal pattern starting from the origin of the coordinate system, and to use one of the components in the final position vector of the iterative process as the difference mutation operator.

[0078] In this embodiment, a two-dimensional fractal pattern is generated starting from the origin using an Iterated Function System (IFS). Specifically, a set of contracting affine transformations containing scaling, rotation, and translation parameters is defined, and a corresponding selection probability is assigned to each transformation. Taking the origin (0,0) as the initial iteration point, a contracting affine transformation is randomly selected from the set according to the above selection probability to perform a transformation operation on the current iteration point, resulting in a new iteration point. This random selection and transformation iteration process is repeated until the preset maximum number of iterations is reached and then terminated. Then, one of the components in the final position vector of the iteration is used as a difference mutation operator.

[0079] Step S2.6.2.3: Select three distinct individuals from the individuals to be mutated, i.e., the first individual. The second body And the third body ;

[0080] Step S2.6.2.4: Select two components from the final coordinates with equal probability, and combine them with the first volume... The second body And the third body , generating a mutated individual v.

[0081] The equal probability selection of two components in the final coordinates, combined with the first volume The second body And the third body Generate a mutated individual v, calculated as follows:

[0082] ;

[0083] Where v represents the mutated individual, x represents the final x-value of the individual to be mutated, and y represents the final x-value of the individual to be mutated. For the first body, For the second body, For the third entity.

[0084] Step S2.6.3: Perform a crossover operation on the mutated individuals to generate offspring individuals after the crossover operation;

[0085] In this embodiment, after generating the mutation vector, the offspring individual is generated by performing a crossover operation between the parent individual and its corresponding mutation vector.

[0086] Step S2.6.4: If the parent individual dominates the child individual, the parent individual is retained and the child individual is eliminated; if the child individual dominates the parent individual, the child individual is retained; if the parent individual does not dominate the child individual and the child individual does not dominate the parent individual, both the parent individual and the child individual are retained.

[0087] To select the dominant individual from the parent and offspring generations that can advance to the next generation of evolution, the algorithm compares the parent individual and its corresponding offspring one by one. "Dominance" here is based on a strict dominance relationship determination: if the parent individual dominates its offspring (i.e., the parent performs no worse than the offspring on all objective functions, and is strictly better than the offspring on at least one objective function), the parent individual is retained; if the offspring individual dominates the parent individual (i.e., the offspring performs no worse than the parent on all objective functions, and is strictly better than the parent on at least one objective function), the offspring individual is retained; if neither dominates the other (i.e., the parent is better on some objectives, and the offspring is better on others, but neither is no worse than the other on all objectives and is strictly better on at least one objective), both are retained.

[0088] In this embodiment, the selection and retention of individuals is performed. To select dominant individuals from both parent and offspring that can enter the next generation of evolution, the algorithm compares parent individuals and their corresponding offspring one by one. If a parent individual dominates its offspring, the parent individual is retained; if an offspring individual dominates its parent, the offspring is retained; if neither dominates the other, both are retained. To address the issue that the resulting offspring population may exceed a predetermined population size, when the offspring population size exceeds a preset limit, a fast non-dominated sorting based on crowding distance is performed on the offspring population. Then, based on the non-dominated sorting level and crowding probability, individuals are selected for retention, reducing the offspring population size back to the preset size.

[0089] Step S2.6.5: If the current iteration count reaches the maximum iteration count, perform non-dominated sorting and use the first level after sorting as the optimized predicted composition data of the cold-rolled product; if the current iteration count does not reach the maximum iteration count, return to step S2.6.2 and continue the differential mutation operation.

[0090] Finally, determine whether the preset termination condition of the MODE algorithm is met. If it is met, perform another fast non-dominated sort on the final population, and use the first level after sorting as the optimization result of the MODE algorithm. If it is not met, treat the current population as the parent population and continue to perform differential mutation operations.

[0091] In the reverse engineering process, the target properties (yield strength YS, tensile strength TS, elongation EL) required for the cold-rolled product are first set. The F-MODE algorithm uses these performance indicators as objective functions and optimizes the chemical composition using a multi-objective evolutionary algorithm by calling the FCNN quality analysis model, gradually bringing it closer to the target properties. To improve the algorithm's search capability, F-MODE introduces a fractal mutation strategy in the mutation operation, thereby enhancing the efficiency of the exploration and optimization process.

[0092] An improved optimization algorithm is used to continuously update the cold-rolled product composition in the decision space based on the target performance for reverse design. The cold-rolled product quality composition design method is expressed in the form of a multi-objective optimization objective function, the purpose of which is to minimize the difference between the predicted performance and the target performance. The objective function calculation formula in this embodiment is as follows:

[0093] ;

[0094] in, , where YS is the decision variable, representing the chemical composition combination of the cold-rolled product; YS is the yield strength (YS) predicted by the trained FCNN model based on the composition of the cold-rolled product; TS is the tensile strength predicted by the trained FCNN model based on the composition of the cold-rolled product; EL is the elongation predicted by the trained FCNN model based on the composition of the cold-rolled product. , , These are three objective functions in a multi-objective optimization problem, used to measure the difference between "predictive performance" and "objective performance," specifically defined as follows: The absolute difference between the predicted yield strength and the target yield strength is the target value. The goal is to minimize this difference so that the predicted yield strength is close to the target value. The absolute difference between the predicted tensile strength and the target tensile strength is the target value. The goal is to minimize this difference so that the predicted yield strength is close to the target value. The absolute difference between the predicted elongation and the target elongation is minimized to bring the predicted elongation closer to the target value. This difference guides the improved optimization algorithm to search for the optimal cold-rolled product composition in the decision space. The optimization algorithm achieves reverse design of the cold-rolled product composition by minimizing the above objective function value.

[0095] This embodiment proposes a multi-objective optimization method for the quality composition of cold-rolled products based on data analysis, using actual production data from a steel plant. The method analyzes the chemical composition of cold-rolled products using an FCNN model, treating the composition as a decision variable and the difference between predicted and target performance as the objective function. An improved MODE algorithm (i.e., the F-MODE algorithm, which incorporates a fractal mutation strategy to introduce a multi-objective differential evolution algorithm for reverse optimization of the corresponding relationships) is then used to optimize the composition. Comparison with actual production data, as shown in Tables 2 and 3, demonstrates that this method exhibits good practicality and stability, effectively optimizing the quality composition of cold-rolled products.

[0096] Table 2. Mean and standard deviation of the absolute values ​​of the deviations between the designed composition and the actual composition of cold-rolled products.

[0097] F-MODE mean / standard deviation F-MODE mean / standard deviation F-MODE mean / standard deviation C <![CDATA[2.12e-02 1.23e-02 ]]> Ni <![CDATA[4.19e-03 2.58e-03 ]]> Alt <![CDATA[9.91e-03 8.24e-03 ]]> Mn <![CDATA[1.17e-01 7.04e-02 ]]> Cu <![CDATA[1.15e-02 9.07e-03 ]]> Nb <![CDATA[8.32e-04 4.52e-04 ]]> S <![CDATA[1.27e-02 7.19e-03 ]]> Mo <![CDATA[2.68e-03 1.91e-03 ]]> N <![CDATA[3.04e-03 1.98e-03 ]]> P <![CDATA[7.71e-03 5.14e-03 ]]> V <![CDATA[1.33e-03 6.43e-04 ]]> As <![CDATA[1.97e-03 1.86e-03 ]]> Si <![CDATA[9.65e-03 8.37e-03 ]]> Ti <![CDATA[4.13e-02 2.63e-02 ]]> Ca <![CDATA[1.60e-03 8.95e-04 ]]> Cr <![CDATA[1.18e-02 9.64e-03 ]]> Als <![CDATA[1.08e-02 7.49e-03 ]]> B <![CDATA[3.74e-04 2.78e-04 ]]>

[0098] Table 3. Mean and standard deviation of the absolute values ​​of the deviations between the designed product performance indicators and the target values.

[0099] YS / MPa TS / MPA EL F-MODE mean / standard deviation <![CDATA[5.19e-02 7.86e-03 ]]> <![CDATA[1.09e-01 2.32e-01 ]]> <![CDATA[2.48e-02 3.87e-02 ]]>

[0100] Example 2:

[0101] This embodiment proposes an electronic device, including: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the aforementioned multi-objective optimization method for the quality composition of cold-rolled products based on data parsing.

[0102] The electronic device may be a mobile phone, computer, or tablet computer, etc., and includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements a multi-objective optimization method for the quality composition of cold-rolled products based on data analysis, as described in the embodiments. It is understood that the electronic device may also include input / output (I / O) interfaces and communication components.

[0103] The processor is used to execute all or part of the steps in the multi-objective optimization method for quality composition of cold-rolled products based on data parsing as described in the above embodiments. The memory is used to store various types of data, which may include, for example, instructions for any application or method in the electronic device, as well as application-related data.

[0104] The processor can be implemented as an Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor, or other electronic components, and is used to execute the multi-objective optimization method for the quality composition of cold-rolled products based on data analysis described in the above embodiments.

[0105] Example 3:

[0106] This embodiment proposes a computer-readable storage medium that stores executable instructions. When these instructions are executed, if they are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0107] The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the multi-objective optimization method for quality composition of cold-rolled products based on data analysis as described in the various embodiments of this application.

[0108] The aforementioned storage media include: flash memory, hard disk, multimedia card, card-type memory (e.g., SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR) memory, random access memory (RAM), static random-access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, disk, optical disk, server, APP (Application) application store, and other media capable of storing program verification codes, on which computer programs are stored. When the computer programs are executed by the processor, they can implement the various steps of the above-described method for suppressing turbulence disturbances in a magnetic levitation natural gas differential pressure generator.

[0109] Example 4:

[0110] This embodiment proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned multi-objective optimization method for the quality composition of cold-rolled products based on data analysis.

[0111] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a computer program product.

[0112] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0113] The scope of protection of this application is not limited to the embodiments described above. Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from the scope and spirit of this disclosure. If such modifications and variations fall within the scope of equivalent technology of this disclosure, then the intent of this disclosure also includes such modifications and variations.

Claims

1. A multi-objective optimization method for the quality composition of cold-rolled products based on data analysis, characterized in that, include: Obtain multiple performance indicators of the cold-rolled product quality to be analyzed; Multiple performance indicators of the cold-rolled product quality to be analyzed are input into the quality analysis and multi-objective optimization model to obtain the composition data of the cold-rolled product corresponding to the multiple indicators. The quality analysis and multi-objective optimization model is established by: constructing the correspondence between the multiple indicators of the cold-rolled product quality and the composition data of the cold-rolled product based on historical data of the cold-rolling production process using a fully connected neural network; and using a fractal mutation strategy to introduce a multi-objective differential evolution algorithm to perform reverse optimization on the correspondence to obtain the quality analysis and multi-objective optimization model.

2. The multi-objective optimization method for quality composition of cold-rolled products based on data analysis according to claim 1, characterized in that, The method, based on historical data from the cold rolling production process, employs a fully connected neural network to construct the correspondence between multiple indicators of cold-rolled product quality and the composition data of cold-rolled products, including: Obtain historical production data for each batch during the cold rolling process and preprocess the historical production data; Feature extraction was performed on the preprocessed historical production data, and the extracted features were correlated with multiple performance indicators of cold-rolled products. A fully connected neural network was used to construct the correspondence between multiple indicators of cold-rolled product quality and the extracted features.

3. The multi-objective optimization method for quality composition of cold-rolled products based on data analysis according to claim 1, characterized in that, The aforementioned performance indicators include: yield strength, tensile strength, and elongation of cold-rolled products.

4. The multi-objective optimization method for quality composition of cold-rolled products based on data analysis according to claim 1, characterized in that, The step of employing a fractal mutation strategy to introduce a multi-objective differential evolution algorithm to perform reverse optimization of the correspondence includes: Step S2.4: Select N different combinations of performance indicators for yield strength, tensile strength and elongation of cold-rolled products from the historical production data of the cold rolling production process, and record the composition data of the actual cold-rolled products corresponding to each combination of performance indicators. Step S2.5: Input each set of performance indexes into a fully connected neural network to obtain the predicted composition data of the cold-rolled product; Step S2.6: Using the predicted composition data of the cold-rolled product as individuals in the population, a multi-objective differential evolution algorithm is introduced using a fractal mutation strategy to optimize the individuals in the population, thereby obtaining the optimized predicted composition data of the cold-rolled product. Step S2.7: Calculate the absolute deviation between the optimized predicted composition data of the cold-rolled product and the actual composition data of the cold-rolled product; Step S2.8: If the absolute deviation is greater than the preset threshold, change the predicted composition data of the cold-rolled product and return to step S2.6 to re-optimize; Step S2.9: If the absolute deviation is less than or equal to the preset threshold, stop the iteration and output the optimized predicted composition data of the cold-rolled product.

5. The multi-objective optimization method for quality composition of cold-rolled products based on data analysis according to claim 4, characterized in that, The process involves using the predicted composition data of cold-rolled products as individuals in a population, employing a fractal mutation strategy to introduce a multi-objective differential evolution algorithm to optimize the individuals in the population, resulting in optimized predicted composition data of the cold-rolled products, including: Step S2.6.1: Initialize the individuals in the population by using the predicted composition data of the cold-rolled product as individuals in the population; Step S2.6.2: Use a multi-objective differential evolution algorithm with fractal mutation strategy to perform differential mutation operation on the initialized individuals to obtain mutated individuals; Step S2.6.3: Perform a crossover operation on the mutated individuals to generate offspring individuals after the crossover operation; Step S2.6.4: If the parent individual dominates the offspring individual it produces, the parent individual will be retained; if the offspring individual dominates the parent individual, the offspring individual will be retained; if neither dominates the other, both will be retained. If the parent individual dominates the offspring individual, the parent individual will be retained and the offspring individual will be eliminated. If the offspring individual dominates the parent individual, the offspring individual will be retained; if neither the parent individual nor the offspring individual dominates the parent individual, both the parent individual and the offspring individual will be retained. Step S2.6.5: If the current iteration count reaches the maximum iteration count, perform non-dominated sorting and use the first level after sorting as the optimized predicted composition data of the cold-rolled product; if the current iteration count does not reach the maximum iteration count, return to step S2.6.2 and continue the differential mutation operation.

6. The multi-objective optimization method for quality composition of cold-rolled products based on data analysis according to claim 5, characterized in that, The multi-objective differential evolution algorithm employing a fractal mutation strategy performs differential mutation operations on the initialized individuals, including: Step S2.6.2.1: Randomly generate the maximum number of iterations from the interval [preset lower limit of maximum number of iterations, preset upper limit of maximum number of iterations]; Step S2.6.2.2: Take the initialized individual as the individual to be mutated, and use the iterative function system to generate the final coordinates of the individual to be mutated starting from the origin. ; Step S2.6.2.3: Select three distinct individuals from the individuals to be mutated, i.e., the first individual. The second body And the third body ; Step S2.6.2.4: Select two components from the final coordinates with equal probability, and combine them with the first volume... The second body And the third body , generating a mutated individual v.

7. The multi-objective optimization method for quality composition of cold-rolled products based on data analysis according to claim 6, characterized in that, The equal probability selection of two components in the final coordinates, combined with the first volume The second body And the third body Generate a mutated individual v, calculated as follows: ； Where v represents the mutated individual, x represents the final x-value of the individual to be mutated, and y represents the final x-value of the individual to be mutated. For the first body, For the second body, For the third entity.

8. An electronic device, characterized in that, include: One or more processors, and a memory for storing instructions that, when executed by the one or more processors, cause the one or more processors to perform a multi-objective optimization method for the quality composition of cold-rolled products based on data analysis, as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed, cause the processor to perform a multi-objective optimization method for the quality composition of cold-rolled products based on data analysis, as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by the processor, they implement the multi-objective optimization method for the quality composition of cold-rolled products based on data analysis as described in any one of claims 1 to 7.

Images