Cold-rolled material quality performance interval forecasting method based on multi-objective evolutionary learning

Through multi-objective evolutionary learning and deep principal component analysis technology, the Gaussian process regression model is optimized, and the problem of high time complexity and kernel function setting dependence of the Gaussian process regression model under large sample data is solved, and the quality performance of cold rolled materials is achieved is achieved. forecast.

CN119067490AActive Publication Date: 2024-12-03NORTHEASTERN UNIV CHINA

Patent Information

Application Number
CN202411086285.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-12-03
Estimated Expiration
2044-08-08

AI Technical Summary

Technical Problem

The existing Gaussian process regression model has high time complexity when processing large sample data and relies on manual setting of kernel functions, resulting in slow training speed and limited prediction accuracy.

Method used

The multi-objective evolutionary learning method is adopted, combined with deep principal component analysis technology and multi-objective differential evolution algorithm, and the hyperparameter and feature extraction of the Gaussian process regression model is optimized to construct a cold-rolled material quality performance interval forecast model.

Benefits of technology

It improves prediction accuracy, reduces time complexity, and reduces dependence on manually set kernel functions, and is suitable for large sample data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067490A_ABST
    Figure CN119067490A_ABST
Patent Text Reader

Abstract

The invention provides a cold-rolled material quality performance interval forecasting method based on multi-objective evolutionary learning, and relates to the technical field of cold-rolled material quality performance forecasting. The method comprises the following steps: firstly, collecting chemical components, process parameters and corresponding quality performance data in the production process of the cold-rolled material to obtain an original sample data set, preprocessing the sample data and dividing the sample data into a training set and a test set; then, a multi-target differential evolution algorithm is adopted to obtain a hyper-parameter set required for constructing a quality performance forecasting model, and an optimal hyper-parameter combination is selected from Pareto frontier based on a preference knee solution to serve as a parameter combination for constructing a Gaussian regression model; and finally, constructing a final cold-rolled material quality performance interval forecasting model based on a Gaussian process regression model framework and the selected kernel function. According to the method, the obtained multi-objective evolutionary learning model has relatively high prediction accuracy and relatively low time complexity.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of cold-rolled material quality performance prediction, and in particular to a cold-rolled material quality performance interval prediction method based on multi-objective evolutionary learning. Background Art

[0002] Cold-rolled materials are one of the most important and widely used materials in industrial production and people's lives, with the characteristics of large-scale production, easy processing, diverse and reliable performance, low price, convenient use and easy recycling. In the processing of cold-rolled materials, strength and shaping are the two most important performance indicators. Among them, strength determines the upper limit of the cold-rolled material's tolerance, while shaping determines the maximum deformation that the cold-rolled material can undergo. Therefore, correctly predicting the performance of cold-rolled materials is an important means to reasonably set the cold-rolling processing technology.

[0003] The quality of cold-rolled materials mainly depends on the chemical composition and process parameters. Generally speaking, the chemical composition of cold-rolled materials includes more than 10 alloying elements in addition to the basic iron and carbon elements. The interaction between alloying elements and iron, carbon and alloying elements changes the phase change process and organization of cold-rolled materials, giving cold-rolled materials many excellent properties. In addition, the processing parameters of cold-rolled materials are also very complex. Taking the heat treatment of cold-rolled materials as an example, the four most commonly used heat treatment methods include annealing, normalizing, quenching and tempering. In order to obtain the optimal microstructure, when using these heat treatment methods, the heating temperature, heating time and many other process parameters must be studied and set. Therefore, in the production process of cold-rolled materials, there will be a large number of combinations of different chemical compositions and process parameters. There are still many difficulties in analyzing these combinations and predicting the properties of steel materials using traditional methods.

[0004] In recent years, with the rapid development of computer technology, the use of machine learning methods to predict the relationship between macroscopic features and microscopic performance has become a research hotspot. The basic principle of machine learning is to allow the computer system to discover patterns and laws from data through learning, and then use these patterns and laws to make predictions or decisions. Among machine learning algorithms, Gaussian process regression has gradually gained the attention of domestic and foreign scholars due to its strict statistical learning theory foundation, good adaptability to complex problems such as small samples and nonlinearity, and strong generalization ability, and has been successfully applied in many fields. Chinese patent CN116070924A provides a tunnel support scheme decision method and system based on Gaussian process regression. The invention can quickly find the best support scheme based on a small sample of support schemes. Chinese patent CN115098928B provides a shield tunnel risk assessment method based on an improved Gaussian process metamodel. The invention uses search functions and learning functions to find the best training points, improves the Gaussian process metamodel, and makes the algorithm more suitable for the analysis of nonlinear functions with potential tail risks. Chinese patent CN112287461B provides a method for modeling a braking system of a car driving simulator based on Gaussian process regression. The invention is based on the actual vehicle experimental data of the braking system and uses the Gaussian process regression modeling method to obtain a high-confidence model of the braking system of the car driving simulator, thereby reducing the dependence of the car driving system on the real braking system hardware and software, and reducing the development and use costs. Although Gaussian process regression performs well in prediction problems in many fields, there are still some problems. First, its time complexity reaches the cube of the number of samples, and the number of sample features also seriously affects its training speed, which is not suitable for large sample data; second, the effect of the Gaussian process regression model depends on the setting of the kernel function, but there is currently no unified theoretical support for how to construct the kernel function and set the hyperparameters of the kernel function. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method for predicting the quality performance interval of cold-rolled materials based on multi-objective evolutionary learning in view of the deficiencies of the above-mentioned prior art, so as to realize the prediction of the quality performance of cold-rolled materials.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for predicting the quality performance interval of cold-rolled materials based on multi-objective evolutionary learning, comprising the following steps:

[0007] Step 1: Collect chemical composition, process parameters and corresponding quality performance data during the cold-rolled material production process to obtain the original sample data set;

[0008] The original sample data set includes n sample data, each sample contains m features and 3 cold-rolled material quality performance indicators, namely yield strength, tensile strength and elongation;

[0009] Step 2: Preprocess the original sample data set and divide it into training set and test set;

[0010] Step 2.1: Process the missing data in the original data sample set and remove the samples containing missing data;

[0011] Step 2.2: Standardize the original sample data set so that the sample data satisfies the standard normal distribution with a mean of 0 and a variance of 1;

[0012] Step 2.3: Divide the data set obtained after the standardization process in step 2.2 into a training data set and a test data set according to a preset ratio;

[0013] Step 3: Determine the kernel function of the Gaussian process regression model;

[0014] The kernel function used in the Gaussian process regression model is the sum of the Matern kernel function and the Gaussian white noise function;

[0015] Step 4: Use a multi-objective differential evolution algorithm to obtain the hyperparameter set required to build a cold-rolled material quality performance prediction model;

[0016] Step 4.1: Set the target dimension of deep principal component analysis feature extraction and the upper and lower limits of the hyperparameters of the Gaussian process kernel function;

[0017] Step 4.2: According to the preset population size, an initial population for pruning the multi-objective differential evolution algorithm is obtained, and the initial population is a hyperparameter that needs to be set in the multi-objective differential evolution algorithm;

[0018] Step 4.3: Perform feature extraction and Gaussian process regression on the cold rolling production feature data in the training data set according to the individual information of the initial population, and record the average absolute error of the corresponding yield strength, tensile strength and elongation prediction on the test data set;

[0019] Step 4.4: Use the multi-objective differential evolution optimization algorithm to perform evolutionary operations, and the operation process is differential mutation, crossover and selection of retained individuals;

[0020] Step 4.4.1: Obtain the mutation operator used in this evolution according to the preset initial mutation operator and the current evolutionary generation, and use the operator to perform differential mutation on the individuals in the parent population to obtain the mutant sub-population;

[0021] The differential mutation strategy expression used in the differential mutation process is as follows:

[0022]

[0023] in, represents the i-th differential mutation vector generated in the t-th iteration, represents three different individuals randomly selected from the population, r1, r2 and r3 are in the interval [0,λ size ] are random integers that are not equal to i and are mutually different, λ size is the size of the population, F is the mutation operator;

[0024] And adopt an adaptive mutation operator strategy, the formula is as follows:

[0025]

[0026]

[0027] In the formula, F0 represents the initial mutation operator, G prune represents the maximum evolutionary generation, G represents the current evolutionary generation;

[0028] Step 4.4.2: Perform a crossover operation based on the parent population, the mutant sub-population, and the preset crossover probability to obtain a crossover sub-population;

[0029] After the mutation vector is generated, the parent individual needs to be selected with its corresponding mutation vector to generate the offspring individual. The formula is as follows:

[0030]

[0031] In the formula, as well as Respectively represent the values ​​of the i-th differential mutation vector, parent vector, and j-th position of the child vector generated in the t-th iteration, Cr is the crossover probability, and sn is a random integer selected from the crossover bit set;

[0032] Step 4.4.3: Perform feature extraction and Gaussian process regression on the cold rolling production feature data in the test data set based on the cross sub-population individual information, and record the average absolute error of the prediction of yield strength, tensile strength and elongation on the test data set;

[0033] Step 4.4.4: Compare the individuals in the parent population with their corresponding individuals in the crossover subpopulation. If one dominates the other, record the dominant individual; otherwise, record both individuals.

[0034] Step 4.4.5: After the comparison is completed, return the recorded individual set;

[0035] Step 4.4.6: Compare the recorded individual set with the preset population size. If the two sizes are the same, the recorded individual is used as the parent population of the next generation. Otherwise, select the individual as the parent population of the next generation from the recorded individual set according to the Pareto rank and crowding degree.

[0036] Step 4.5: If the current evolutionary generation is less than the preset evolutionary generation, return to step 4.4, otherwise, return to the final population;

[0037] Step 5: Perform non-dominated sorting on the final population to obtain its Pareto front, and then select the preferred knee solution in the front;

[0038] Step 6: Based on the individual information of the preferred knee solution in the final population, feature extraction and Gaussian process regression are performed on the cold rolling production feature data in the training data set to obtain the final cold rolling material quality performance interval prediction model;

[0039] Step 6.1: Use deep principal component analysis technology to extract features from the chemical composition of cold-rolled materials and process parameters of cold-rolled processes in the training data set; the structures of the first and second layers of the deep principal component analysis technology are exactly the same, and each layer first performs zero-phase principal component analysis whitening on the input data of the layer, and then uses principal component analysis to extract features; the target dimension of the principal component analysis feature extraction is obtained by a multi-objective differential evolution optimization algorithm;

[0040] Step 6.1.1: Perform zero-phase principal component analysis whitening on the chemical composition of cold-rolled materials and process parameters of cold-rolled process in the training data set, and then perform principal component analysis to obtain shallow information of the training data;

[0041] Step 6.1.2: Perform zero-phase principal component analysis whitening on the shallow information obtained in step 6.1.1, and then perform principal component analysis to obtain deep information;

[0042] Step 6.1.3: Concatenate the feature extraction results obtained in step 6.1.1 and step 6.1.2 as the input of the Gaussian process regression model;

[0043] Step 6.2: Use the Gaussian process regression model to predict the yield strength, tensile strength and elongation of the cold rolled material and provide a prediction error band;

[0044] The Gaussian process regression model is a set of any finite random variables with a joint Gaussian distribution, which uses different Gaussian distributions to represent each prediction output, and the mean of these Gaussian distributions is used as the prediction value of the Gaussian process regression model for different cold-rolled material properties; in addition, the mean and standard deviation of these Gaussian distributions are used to construct a prediction error band for each cold-rolled material property, and the expression is as follows:

[0045] b=μ±α×σ (5)

[0046] Among them, b is the upper and lower boundary values ​​of the prediction error band, μ is the mean of the Gaussian distribution, σ is the standard deviation of the Gaussian distribution, and α is a constant.

[0047] The method of the present invention obtains the chemical composition, process parameters and corresponding quality performance data of the cold-rolled material production process, obtains the hyperparameters required by the cold-rolled material quality performance interval prediction model through a multi-objective differential evolution algorithm, extracts features of the cold-rolled composition and process data and performs Gaussian process regression according to the individual information of the preferred knee solution in the final population, and obtains the final cold-rolled material quality performance interval prediction model. Compared with relying on manual pre-definition, the obtained Gaussian process regression model has higher prediction accuracy and lower time complexity.

[0048] The beneficial effects of adopting the above technical solution are: the method for cold-rolled material quality performance interval prediction based on multi-objective evolutionary learning provided by the present invention is aimed at the fact that the setting of the kernel function of the current Gaussian process regression model largely relies on manual pre-definition, and has a high degree of dependence on the professional knowledge level and practical experience of the technicians. In addition, Gaussian process regression involves a large number of matrix inversion operations during the training process, and it takes a long time when facing large samples and high-dimensional data. The advantages of multi-objective differential evolution algorithm, deep principal component analysis technology and Gaussian process regression technology are integrated, and deep principal component analysis technology is used to extract features from large sample data sets, and the multi-objective differential evolution algorithm is used to simultaneously optimize the feature extraction target dimension and the Gaussian process kernel function hyperparameters, so that the obtained multi-objective evolutionary learning model has a higher prediction accuracy and lower time complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A flow chart of a method for predicting cold-rolled material quality performance intervals based on multi-objective evolutionary learning provided by an embodiment of the present invention;

[0050] Figure 2 A schematic diagram of reduced-dimensional Gaussian process regression provided by an embodiment of the present invention;

[0051] Figure 3 A multi-objective differential evolution optimization flow chart provided by an embodiment of the present invention;

[0052] Figure 4 The final population graph of differential evolution provided by the embodiment;

[0053] Figure 5The predicted error band diagrams of the quality performance indicators of cold-rolled materials provided in the embodiments, wherein (a) is the predicted error band diagram of the yield strength indicator, (b) is the predicted error band diagram of the tensile strength indicator, and (c) is the predicted error band diagram of the elongation indicator. DETAILED DESCRIPTION

[0054] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0055] In this implementation, the cold-rolled material quality performance interval prediction method based on multi-objective evolutionary learning uses deep principal component analysis technology and multi-objective differential evolution optimization algorithm to extract data features and optimize model parameters to generate a reduced-dimensional Gaussian process regression model with high prediction accuracy, and can also provide a prediction error band.

[0056] In this implementation, a cold-rolled material quality performance interval prediction method based on multi-objective evolutionary learning is proposed, such as Figure 1 As shown, the following steps are included:

[0057] Step 1: Collect chemical composition, process parameters and corresponding quality performance data during the cold-rolled material production process to obtain the original sample data set.

[0058] The original sample data set collected in this embodiment includes 65,288 data, each of which contains 27 features and 3 cold-rolled material quality performance indicators, namely yield strength, tensile strength and elongation;

[0059] Step 2: Preprocess the original sample data set and divide it into training set and test set;

[0060] Step 2.1: Process the missing data in the original data sample set and remove the samples containing missing data.

[0061] In this embodiment, 2151 pieces of data in the original sample data set contain missing values. After removing these sample data, the remaining 63137 pieces of data are used for the cold-rolled material quality performance prediction in this embodiment;

[0062] Step 2.2: Standardize the original sample data set so that the sample data satisfies the standard normal distribution with a mean of 0 and a variance of 1;

[0063] Step 2.3: Divide the data set obtained after the standardization process in step 2.2 into a training data set and a test data set according to a preset ratio.

[0064] This embodiment adopts a random selection method, where 80% of the data input into the Gaussian process regression model each time is used for training, and the remaining 20% ​​is used for testing;

[0065] Step 3: Determine the kernel function of the Gaussian process regression model;

[0066] In this embodiment, the kernel function used by the Gaussian process regression model is the sum of the Matern kernel function and the Gaussian white noise function. The mathematical expression of the Matern kernel function is shown in Formula 1. In this embodiment, v=1.5 is selected, and l is the hyperparameter to be optimized.

[0067] The expression of the Gaussian white noise kernel function is shown in Formula 2, where n is the hyperparameter to be optimized.

[0068]

[0069] k(x i ,x j )=n if x i = = x j else 0 (2)

[0070] Step 4: Use a multi-objective differential evolution algorithm to obtain the hyperparameter set required to build a cold-rolled material quality performance prediction model;

[0071] Step 4.1: Set the target dimension of deep principal component analysis feature extraction and the upper and lower limits of the hyperparameters of the Gaussian process kernel function;

[0072] In this embodiment, the target dimension range of the first principal component analysis dimensionality reduction in the deep principal component analysis process is set to [6, 27], the target dimension range of the second principal component analysis dimensionality reduction is set to [5, 27], the value range of the hyperparameter l in the Matern kernel function is set to [1, 10], and the value range of the hyperparameter n in the Gaussian white noise function is set to [0, 1];

[0073] Step 4.2: According to the preset population size, an initial population for pruning the multi-objective differential evolution algorithm is obtained. The initial population is a hyperparameter that needs to be set in the multi-objective differential evolution algorithm. In this embodiment, the preset population size is 30;

[0074] Step 4.3: Perform feature extraction and Gaussian process regression on the cold rolling production feature data in the training data set according to the individual information of the initial population, and record the average absolute error of the corresponding yield strength, tensile strength and elongation prediction on the test data set;

[0075] Step 4.4: Use the multi-objective differential evolution optimization algorithm to perform evolutionary operations, and the operation process is differential mutation, crossover and selection of retained individuals;

[0076] Step 4.4.1: Obtain the mutation operator used in this evolution according to the preset initial mutation operator and the current evolutionary generation, and use the operator to perform differential mutation on the individuals in the parent population to obtain the mutant sub-population;

[0077] In this embodiment, the differential mutation strategy expression used is as follows:

[0078]

[0079] in, represents the i-th differential mutation vector generated in the t-th iteration, represents three different individuals randomly selected from the population, r1, r2 and r3 are in the interval [0,λ size ] are random integers that are not equal to i and are mutually different, λ size is the size of the population, and F is the mutation operator.

[0080] In the search process of the basic differential evolution algorithm, if the mutation operator F takes a large value, the algorithm search efficiency is low, and the accuracy of the global optimal solution obtained is low. If the mutation operator F takes a small value, the population diversity is reduced, and the problem of "premature maturity" is prone to occur. This embodiment adopts a strategy of an adaptive mutation operator, and its formula is as follows:

[0081]

[0082]

[0083] In the formula, F0 represents the initial mutation operator, which is set to 0.4 in this embodiment, G prune represents the maximum evolutionary generation, and G represents the current evolutionary generation.

[0084] At the beginning of the differential evolution algorithm, the adaptive mutation operator has a large value, which maintains individual diversity in the early stage and avoids the occurrence of "premature" phenomenon. With the continuous iteration of the algorithm, the mutation operator gradually decreases, and in the later stage the mutation operator is close to F0, retaining good information, avoiding the destruction of the optimal solution, and increasing the probability of searching for the global optimum.

[0085] Step 4.4.2: Perform a crossover operation based on the parent population, the mutant sub-population, and the preset crossover probability to obtain a crossover sub-population;

[0086] After the mutation vector is generated, the parent individual needs to be selected with its corresponding mutation vector to generate the offspring individual. The formula is as follows:

[0087]

[0088] In the formula, as well as They respectively represent the values ​​of the i-th differential mutation vector, parent vector and j-th position of the child vector generated in the t-th iteration, Cr is the crossover probability, which is set to 0.1 in this embodiment, and sn is a random integer selected from the crossover bit set.

[0089] By performing the above differential mutation and selection operations, the offspring population used to select and retain individuals is obtained.

[0090] Step 4.4.3: Perform feature extraction and Gaussian process regression on the cold rolling production feature data in the test data set based on the cross sub-population individual information, and record the average absolute error of the prediction of yield strength, tensile strength and elongation on the test data set;

[0091] Step 4.4.4: Compare the individuals in the parent population with their corresponding individuals in the crossover subpopulation. If one dominates the other, record the dominant individual; otherwise, record both individuals.

[0092] Step 4.4.5: After the comparison is completed, return the recorded individual set;

[0093] Step 4.4.6: Compare the recorded individual set with the preset population size. If the two sizes are the same, the recorded individual is used as the parent population of the next generation. Otherwise, select the individual as the parent population of the next generation from the recorded individual set according to the Pareto rank and crowding degree.

[0094] Step 4.5: If the current evolutionary generation is less than the preset evolutionary generation of 60, return to step 4.4, otherwise, return to the final population;

[0095] Step 5: Perform non-dominated sorting on the final population to obtain its Pareto front, and then select the preferred knee solution in the front;

[0096] Step 6: Based on the individual information of the preferred knee solution in the final population, feature extraction and Gaussian process regression are performed on the cold rolling production feature data in the training data set to obtain the final cold rolling material quality performance interval prediction model;

[0097] Step 6.1: Use deep principal component analysis technology to extract features of the chemical composition of cold-rolled materials and process parameters of cold-rolling process in the training data set;

[0098] Step 6.1.1: Perform zero-phase principal component analysis whitening on the chemical composition of cold-rolled materials and process parameters of cold-rolled process in the training data set, and then perform principal component analysis to obtain shallow information of the training data;

[0099] Step 6.1.2: Perform zero-phase principal component analysis whitening on the shallow information obtained in step 6.1.1, and then perform principal component analysis to obtain deep information;

[0100] Step 6.1.3: Concatenate the feature extraction results obtained in step 6.1.1 and step 6.1.2 as the input of the Gaussian process regression model;

[0101] Step 6.2: Use the Gaussian process regression model to predict the yield strength, tensile strength and elongation properties of the cold rolled material and provide the prediction error band.

[0102] In this embodiment, the Gaussian process regression model is as follows: Figure 2 As shown. Figure 2 It can be seen that the structures of the first and second layers of the deep principal component analysis technology are exactly the same. Each layer first performs zero-phase principal component analysis whitening on the input data of that layer, and then uses principal component analysis to extract features. Zero-phase principal component analysis whitening is a classic data preprocessing algorithm. Whitening is performed to reduce the information redundancy of the original data as much as possible, reduce the correlation between the data, and make all the data have the same variance.

[0103] The specific process of zero-phase principal component analysis whitening is as follows:

[0104] First, the cold rolling production characteristic data including the chemical composition of cold rolling materials and the process parameters of cold rolling are centered, that is, the mean operation is removed, and the mean μ of N data is calculated, where x i Represents a piece of cold rolling production feature data, including multiple features:

[0105]

[0106] Each cold rolling production characteristic data x i Subtract the mean μ:

[0107] x' i =(x i -μ) (8)

[0108] Calculate the covariance matrix M for the centralized cold rolling production characteristic data:

[0109]

[0110] Perform singular value decomposition on the covariance matrix:

[0111] M=UΛA T (10)

[0112] Among them, Λ=diag(λ1,λ2,…,λ n ), λ irepresents the eigenvalues ​​obtained by decomposing the covariance matrix, n is the number of eigenvalues, and V is the eigenvector matrix.

[0113] Calculate the zero-phase principal component analysis whitening matrix P w :

[0114] P w =VΛ'- 1 / 2 U T (11)

[0115] Among them, Λ'=diag(λ1+ε,λ2+ε,...,λ n +ε), where ε represents a very small parameter. Here, adding a very small positive integer ε to each eigenvector can eliminate the influence of some noise in the process of obtaining data information to a certain extent.

[0116] After obtaining the whitening matrix, for any given cold rolling process characteristic data, the result after zero-phase principal component analysis whitening is:

[0117] x'=P w T x (12)

[0118] In the whole zero-phase principal component analysis whitening process, no dimensionality reduction is performed on the data. After whitening, the correlation between the data is reduced, the variance is 1, and the redundancy is reduced.

[0119] Next, principal component analysis is performed on the whitened data set to extract features. The target dimension of the principal component analysis feature extraction is obtained through a multi-objective differential evolution optimization algorithm.

[0120] The deep principal component analysis algorithm cascades the first-layer feature F1 and the second-layer feature F2 to obtain the final feature F, which is used to model the Gaussian process regression model. The final feature F can be expressed as:

[0121]

[0122] The Gaussian process regression model is a set of any finite random variables with a joint Gaussian distribution. It uses different Gaussian distributions to represent each prediction output. The mean of these Gaussian distributions can be used as the prediction value of the Gaussian process regression model for different cold-rolled material properties. In addition, using the mean and standard deviation of these Gaussian distributions, the prediction error band for each cold-rolled material property can be constructed, and the expression is as follows:

[0123] b=μ±α×σ (14)

[0124] Among them, b is the upper and lower boundary values ​​of the prediction error band, μ is the mean of the Gaussian distribution, σ is the standard deviation of the Gaussian distribution, and α is a constant; in this embodiment, when constructing the prediction error band of the yield strength index and the tensile strength index, α is 3, and when constructing the prediction error band of the elongation index, α is 4.

[0125] The multi-objective differential evolution optimization process is as follows Figure 3 As shown. Before entering the pruning process, the algorithm will encode the Gaussian process regression model. In this embodiment, real number encoding is used to represent population individuals, and the encoding string is as follows:

[0126] Model = [d1, d2, l, n] (15)

[0127] Among them, d1 is the target dimension of the first layer of principal component analysis dimensionality reduction of deep principal component analysis, d2 is the target dimension of the second layer of principal component analysis dimensionality reduction of deep principal component analysis, and l and n are the hyperparameters of the kernel function of the Gaussian process regression model.

[0128] The multi-objective differential evolution optimization algorithm first randomly initializes the encoding string of the population size as the initial parent population, and then cyclically performs differential mutation, crossover, and selection of retained individuals in the preset evolutionary algebra. In the process of selecting retained individuals, the algorithm compares each individual of the parent population in this cycle and the offspring individuals generated after differential mutation and crossover operations. If the parent individual dominates the offspring individual it generates, the parent individual will be retained; if the offspring individual dominates the parent individual, the offspring individual will be retained; if the two do not dominate each other, both will be retained. The offspring population generated in this way may exceed the specified population size. In order to solve this problem, when the offspring population size exceeds the preset, the offspring population is non-dominated sorted and the crowding degree of individuals in each Pareto set is calculated. Then, the retained individuals are selected according to the Pareto set level and crowding degree, and the offspring population size is reduced back to the preset size. Figure 4 This is the Pareto front diagram of the final population of multi-objective differential evolution. The individuals marked with triangles in the figure are the individuals finally selected.

[0129] In this embodiment, the first optimization goal considered by the multi-objective differential evolution optimization algorithm is the first and second level principal component analysis dimension reduction target dimensions. The second optimization goal is the sum of the average absolute error of the predicted yield strength, the average absolute error of the predicted tensile strength, and the average absolute error of the predicted elongation by the reduced dimensionality Gaussian process regression model. In order to unify the magnitude of the three average absolute errors, this embodiment increases the average absolute error of the predicted elongation by 10 times when adding them.

[0130] In this embodiment, the cold-rolled material quality performance interval prediction method based on multi-objective evolutionary learning of the present invention is used to simultaneously model and predict the three quality performance indicators in the data set. When modeling and predicting each quality performance indicator each time, 32,000 data are randomly selected from the data set, and a total of 5 predictions are performed to record the optimal value and average value of each regression performance evaluation indicator.

[0131] The regression performance evaluation index used in this embodiment includes the evaluation coefficient (R 2 ), mean absolute error (MAE) and root mean square error (RMSE), the formulas are as follows:

[0132]

[0133] Among them, y i represents the true value of the ith observation, represents the predicted value of the ith observation, represents the average of the true values ​​of all observations, and n represents the number of samples.

[0134] In this embodiment, the prediction results of the three quality performance indicators in the data set are shown in Table 1;

[0135] Table 1 Prediction results of three quality performance indicators

[0136]

[0137]

[0138] In this embodiment, the prediction error band for yield strength is constructed as follows: Figure 5 As shown in (a), the prediction error band for tensile strength is Figure 5 As shown in (b), the prediction error band for elongation is Figure 5 In this embodiment, the maximum hit rate of yield strength is 0.9804, the average hit rate is 0.9553, the maximum hit rate of tensile strength is 0.9918, the average hit rate is 0.9852, and the maximum hit rate of elongation is 0.9538, the average hit rate is 0.8991.

[0139] It can be seen from the examples that the method of the present invention can provide prediction error bands for different performance indicators while more accurately predicting the yield strength, tensile strength and elongation of cold-rolled materials, making it convenient for on-site operators to flexibly set subsequent cold-rolled material production and processing parameters according to the predicted maximum and minimum values ​​of different performance indicators, thereby improving the performance of the final product.

[0140] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A method for cold-rolled material quality performance interval prediction based on multi-objective evolutionary learning, characterized by: The following steps are involved: Step 1: Collect chemical composition, process parameters and corresponding quality performance data during the cold-rolled material production process to obtain the original sample data set; Step 2: Preprocess the original sample data set and divide it into training set and test set; Step 3: Determine the kernel function of the Gaussian process regression model; Step 4: Use a multi-objective differential evolution algorithm to obtain the hyperparameter set required to build a cold-rolled material quality performance prediction model; Step 4.1: Set the target dimension of deep principal component analysis feature extraction and the upper and lower limits of the hyperparameters of the Gaussian process kernel function; Step 4.2: According to the preset population size, an initial population for pruning the multi-objective differential evolution algorithm is obtained, and the initial population is a hyperparameter that needs to be set in the multi-objective differential evolution algorithm; Step 4.3: Perform feature extraction and Gaussian process regression on the cold rolling production feature data in the training data set according to the individual information of the initial population, and record the average absolute error of the corresponding yield strength, tensile strength and elongation prediction on the test data set; Step 4.4: Use the multi-objective differential evolution optimization algorithm to perform evolutionary operations, and the operation process is differential mutation, crossover and selection of retained individuals; Step 4.5: If the current evolutionary generation is less than the preset evolutionary generation, return to step 4.4, otherwise, return to the final population; Step 5: Perform non-dominated sorting on the final population to obtain its Pareto front, and then select the preferred knee solution in the front; Step 6: Based on the individual information of the preferred knee solution in the final population, feature extraction and Gaussian process regression are performed on the cold rolling production feature data in the training data set to obtain the final cold rolling material quality performance interval prediction model; Step 6.1: Use deep principal component analysis technology to extract features from the chemical composition of cold-rolled materials and process parameters of cold-rolled processes in the training data set; the structures of the first and second layers of the deep principal component analysis technology are exactly the same, and each layer first performs zero-phase principal component analysis whitening on the input data of the layer, and then uses principal component analysis to extract features; the target dimension of the principal component analysis feature extraction is obtained by a multi-objective differential evolution optimization algorithm; Step 6.2: Use the Gaussian process regression model to predict the yield strength, tensile strength and elongation properties of the cold rolled material and provide the prediction error band.

2. The method for cold-rolled material quality performance interval prediction based on multi-objective evolutionary learning according to claim 1 is characterized in that: The original sample data set described in step 1 includes n sample data, each of which contains m features and three cold-rolled material quality performance indicators, namely yield strength, tensile strength and elongation.

3. The method for cold-rolled material quality performance interval prediction based on multi-objective evolutionary learning according to claim 2 is characterized in that: The step 2 comprises: Step 2.1: Process the missing data in the original data sample set and remove the samples containing missing data; Step 2.2: Standardize the original sample data set so that the sample data satisfies the standard normal distribution with a mean of 0 and a variance of 1; Step 2.3: Divide the data set obtained after the standardization process in step 2.2 into a training data set and a test data set according to a preset ratio.

4. The method for cold-rolled material quality performance interval prediction based on multi-objective evolutionary learning according to claim 1 is characterized in that: The kernel function used in the Gaussian process regression model in step 3 is the sum of the Matern kernel function and the Gaussian white noise function.

5. The method for cold-rolled material quality performance interval prediction based on multi-objective evolutionary learning according to claim 3 is characterized in that: The step 4.4 comprises: Step 4.4.1: Obtain the mutation operator used in this evolution according to the preset initial mutation operator and the current evolutionary generation, and use the operator to perform differential mutation on the individuals in the parent population to obtain the mutant sub-population; Step 4.4.2: Perform a crossover operation based on the parent population, the mutant sub-population, and the preset crossover probability to obtain a crossover sub-population; Step 4.4.3: Perform feature extraction and Gaussian process regression on the cold rolling production feature data in the test data set based on the cross sub-population individual information, and record the average absolute error of the prediction of yield strength, tensile strength and elongation on the test data set; Step 4.4.4: Compare the individuals in the parent population with their corresponding individuals in the crossover subpopulation. If one dominates the other, record the dominant individual; otherwise, record both individuals. Step 4.4.5: After the comparison is completed, return the recorded individual set; Step 4.4.6: Compare the recorded individual set with the preset population size. If the two are of the same size, the recorded individual is used as the parent population of the next generation. Otherwise, the individual as the parent population of the next generation is selected from the recorded individual set based on the Pareto rank and crowding degree.

6. The method for cold-rolled material quality performance interval prediction based on multi-objective evolutionary learning according to claim 5 is characterized in that: The differential mutation strategy expression used in the differential mutation process described in step 4.4.1 is as follows: in, represents the i-th differential mutation vector generated in the t-th iteration, represents three different individuals randomly selected from the population, r1, r2 and r3 are in the interval [0,λ size ] are random integers that are not equal to i and are mutually different, λ size is the size of the population, F is the mutation operator; And adopt an adaptive mutation operator strategy, the formula is as follows: F=F0×2 λ (3) In the formula, F0 represents the initial mutation operator, G prune represents the maximum evolutionary generation, and G represents the current evolutionary generation.

7. The method for cold-rolled material quality performance interval prediction based on multi-objective evolutionary learning according to claim 6 is characterized in that: After the mutation vector is generated in step 4.4.2, the offspring individuals are generated by performing a selection operation on the parent individuals and the mutation vectors corresponding to the parent individuals. The formula is as follows: In the formula, as well as They represent the values ​​of the i-th differential mutation vector, parent vector, and j-th position of the child vector generated in the t-th iteration, Cr is the crossover probability, and sn is a random integer selected from the crossover bit set.

8. The method for cold-rolled material quality performance interval prediction based on multi-objective evolutionary learning according to claim 7 is characterized in that: The step 6.1 comprises: Step 6.1.1: Perform zero-phase principal component analysis whitening on the chemical composition of cold-rolled materials and process parameters of cold-rolled process in the training data set, and then perform principal component analysis to obtain shallow information of the training data; Step 6.1.2: Perform zero-phase principal component analysis whitening on the shallow information obtained in step 6.1.1, and then perform principal component analysis to obtain deep information; Step 6.1.3: Concatenate the feature extraction results obtained in step 6.1.1 and step 6.1.2 as the input of the Gaussian process regression model.

9. The method for cold-rolled material quality performance interval prediction based on multi-objective evolutionary learning according to claim 8, characterized in that: The Gaussian process regression model is a set of any finite random variables with a joint Gaussian distribution, which uses different Gaussian distributions to represent each prediction output, and the mean of these Gaussian distributions is used as the prediction value of the Gaussian process regression model for different cold-rolled material properties; in addition, the mean and standard deviation of these Gaussian distributions are used to construct a prediction error band for each cold-rolled material property, and the expression is as follows: b=μ±α×σ (5) Among them, b is the upper and lower boundary values ​​of the prediction error band, μ is the mean of the Gaussian distribution, σ is the standard deviation of the Gaussian distribution, and α is a constant.

Citation Information

Patent Citations

  • Process for the production of fine chemicals

    CN101421412A

  • Evolutionary algorithm for sensor position optimization during space-time modeling of distributed parameter system based on two codes

    CN112861294A

  • Material optimal correlation relation obtaining method based on genetic algorithm and application

    CN116486969A

  • Machine Learning to Accelerate Design of Energetic Materials

    US20220067249A1

Cited By

  • Cold pier steel ball quality prediction and optimization method based on big data analysis

    CN120013337A

  • Continuous annealing strip steel quality multi-index prediction method based on cross-scale evolution feature learning

    CN120706638A