A method for building a silicon steel quality prediction model based on meta-heuristic optimization algorithm

Abstract

The present invention relates to the technical field of silicon steel quality prediction, and solves the technical problems in the prior art that the quality of silicon steel is difficult to predict comprehensively, and the prediction results are inaccurate and unreliable. In particular, it relates to a method for constructing a silicon steel quality prediction model based on a meta-heuristic optimization algorithm, and the method comprises the following steps: obtaining silicon steel data of overall silicon steel processing, and constructing a basic silicon steel prediction model based on the silicon steel data, wherein the silicon steel data includes all processes and process parameters of overall silicon steel processing; determining the upper and lower limits of the process parameters according to any silicon steel production process standard, and setting the meta-heuristic optimization algorithm according to the upper and lower limits of the process parameters to initialize the relevant parameters of the particle population. The present invention uses a BP neural network optimized based on a meta-heuristic algorithm to construct a quality prediction model for silicon steel quality prediction, comprehensively considers multiple processes and multiple process parameters, realizes the prediction of product quality of all processes and multiple process parameters, and ensures the accuracy of model prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of silicon steel quality prediction, and in particular to a method for constructing a silicon steel quality prediction model based on a meta-heuristic optimization algorithm. Background Art

[0002] Quality control is crucial in the production process of silicon steel. As one of the core materials in power equipment, silicon steel directly affects the performance and efficiency of the equipment. However, in the complex manufacturing process, the quality problems of silicon steel are often difficult to predict. Problems such as shape exceeding the specified accuracy and surface defects often exist. These problems not only affect the appearance quality of silicon steel products, but also directly affect their magnetic properties and loss characteristics.

[0003] Quality prediction has become a key step in ensuring the quality of silicon steel products. By predicting the quality of silicon steel during the manufacturing process, producers can adjust process parameters in a timely manner, optimize production processes, improve product consistency and performance, and thus improve production efficiency and economic benefits.

[0004] However, there are many challenges in practical applications. The manufacturing process of silicon steel is affected by many factors, such as material removal behavior, internal stress evolution, microstructure evolution, and the complex coupling of mechanical, special and composite energy fields. These factors make the quality prediction of silicon steel products extremely complicated. At present, although there are many quality prediction methods based on machine learning algorithms, they often only consider a single process or fewer process parameters, and cannot fully reflect the factors affecting product quality, resulting in inaccuracy and low reliability of the prediction results.

[0005] Therefore, there is an urgent need for a new algorithm for silicon steel quality prediction that can comprehensively consider multiple processes and multiple process parameters. Based on the current research status, this paper aims to propose a process optimization and quality prediction model based on a meta-heuristic optimization algorithm combined with a BP neural network to solve the above challenges, thereby providing an effective solution for quality control and production efficiency improvement of silicon steel products. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention provides a method for constructing a silicon steel quality prediction model based on a meta-heuristic optimization algorithm, which solves the technical problems in the prior art that silicon steel quality is difficult to predict comprehensively and the prediction results are inaccurate and unreliable.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for constructing a silicon steel quality prediction model based on a meta-heuristic optimization algorithm, the method comprising the following steps:

[0008] S1. Obtain silicon steel data of overall silicon steel processing, and build a basic silicon steel prediction model based on the silicon steel data, wherein the silicon steel data includes all processes and process parameters of overall silicon steel processing;

[0009] S2. Determine the upper and lower limits of process parameters according to any silicon steel production process standard, and set the relevant parameters of the meta-heuristic optimization algorithm for initializing the particle population according to the upper and lower limits of the process parameters;

[0010] S3, determine the position of the current individual particle in the particle population within the feasible area based on the FENR search rule, and calculate the fitness value fitness according to the position of the current individual particle;

[0011] S4, updating the global optimal position according to the fitness value fitness of the current individual particle to obtain the global optimal position, and inputting its position information into the basic silicon steel prediction model to realize the update of process parameters;

[0012] S5. Determine whether the global optimal position is within the process standard range according to the upper and lower limits of the process parameters;

[0013] If so, then end;

[0014] If not, the parameters of the basic silicon steel prediction model are optimized, and then the process returns to step S3.

[0015] Furthermore, in step S1, the specific process includes the following steps:

[0016] S11, record the collected processing data of the entire process, and organize and mark them according to the corresponding relationship between the process and the process parameters;

[0017] S12, storing the collected processing data of the entire process into a database, and using the data storage method of the database to obtain complete silicon steel data;

[0018] S13. Use complete silicon steel data to train the BP neural network to obtain a basic silicon steel prediction model.

[0019] Furthermore, in step S2, the specific process includes the following steps:

[0020] S21. Determine the upper and lower limits of the process parameters according to the production process standard of any silicon steel corresponding to the grade, and use the normalization method to normalize the upper and lower limits of the process parameters to between [-1,1];

[0021] S22, using formula The normalized upper and lower limits of the process parameters are used to generate random particle populations, where represents the position of the jth dimension of the nth population, rand represents a random number between (0,1), ub represents the upper bound of random number generation, i.e., upper bound; lb represents the lower bound, i.e., lower bound;

[0022] S23, initialize the relevant parameters x of the particle population n , x n The population matrix of the population in all dimensions can be depicted, namely:

[0023]

[0024] Where n = 1, 2, ...; N p =1,2,…,dim; j=1,2,…,dim.

[0025] Furthermore, in step S3, the specific process includes the following steps:

[0026] S31, using FENR search rules and related parameters x n+1 Allow particle populations to explore feasible areas more accurately and obtain better positions;

[0027]

[0028] x n+1 =x n -FENR

[0029] Among them, randn represents a normally distributed random number with a mean of 0 and a variance of 1; X w Indicates the worst position; X b Indicates the best position, Indicates the current location;

[0030] S32. Current location Update to get the current position of individual particles The update formula is:

[0031]

[0032] Among them, r 1 and r 2 are distinct integers randomly selected from the population; and They represent the position vectors of two different individuals randomly selected from the population; a and b are preset constants; fitness is the fitness value;

[0033] S33, according to the current position of individual particles Calculate the fitness value fitness in order to select the best solution and use it for further optimization operations.

[0034] Furthermore, in step S33, the fitness value fitness is calculated as:

[0035] fitness=-α×MSE+(1-α)×R 2

[0036]

[0037] Among them, y i is the actual measured value of the i-th sample, i.e., the target value; f(x i ) is the prediction value of the basic silicon steel prediction model for the i-th sample; MSE is the mean prediction error; R 2 is a fraction, indicating the proportion of variance explained by the basic silicon steel prediction model; α is a trade-off between the mean square error and R 2 The parameter of the score contribution has a value range of (0≤α≤1); fitness is the fitness value; N is the total number of samples; is the average value of the actual measured values.

[0038] Furthermore, in step S4, the specific process includes the following steps:

[0039] S41, updating FENR search rules to update the global optimal position to obtain the global optimal position Specifically, the expression for updating the FENR search rule is:

[0040]

[0041] in,

[0042] y w =r 1 ×(Mean(Z n+1 +x n )+r 1 ×Δx)×fitness(Z n+1 )

[0043] y b =r 1 ×(Mean(Z n+1 +x n )-r 1 ×Δx)×fitness(Z n+1 )

[0044]

[0045] In the above formula, Z n+1is the next position calculated based on the current position and the random number; y w and b Use Z n+1 and x n The positions of the two vectors generated; (fitness(y w )-fitness(y b )) is the fitness difference, which is used to weight the FENR search rule, emphasizing that the positions with higher fitness contribute more to the update; r 1 Represents a random number between (0,1); x n are related parameters; X w Indicates the worst position; X b Indicates the best position, Indicates the current location.

[0046] Global optimal position The expression is:

[0047]

[0048] in, and They represent the position vectors of two different individuals randomly selected from the population; a and b are preset constants; To update the current location Get the current position of the individual particle.

[0049] S42, the global optimal position Bring it into the trap avoidance algorithm based on fitness value to get the location information

[0050] S43, the global optimal position Location information It is used as input to the basic silicon steel prediction model to update process parameters.

[0051] Further, in step S42, the location information The calculation formula is:

[0052]

[0053] μ 1 =β×3×rand+(1-β)

[0054] μ 2 =β×rand+(1-β)

[0055] Among them, θ 1 ,θ 2are uniform random numbers between (-1,1) and (-0.5,0.5); μ 1 , μ 2 is a random number; rand is a uniform random number between (0,1); β is the exploration adjustment factor, and its value range is [0,1].

[0056] Furthermore, in step S5, the internal parameters of the basic silicon steel prediction model are optimized, and the specific process includes the following steps:

[0057] S51, extracting the internal parameters of the basic silicon steel prediction model and saving them as a parameter set X and encoding them to obtain binary-coded internal parameters;

[0058] S52, set the number of neurons n in the input layer of the BP neural network input =3, the number of neurons in the hidden layer n hidden =10, the number of neurons in the output layer n output =1, the weight matrix between the input layer and the hidden layer is W1, the weight matrix between the hidden layer and the output layer is W2, and the bias vectors of the hidden layer and the output layer are b1 and b2 respectively;

[0059] S53, decoding the binary-coded internal parameters into real values, and mapping them to obtain optimized process internal parameters, which includes mapping the decoded process parameters to a real value of size n. hidden ×n input Each element position of the weight matrix W1 is mapped to a size of n output ×n hidden Each element position of the weight matrix W2 and the length of the mapping are n hidden and n output Each element position of the bias vectors b1 and b2 is finally obtained to obtain the optimized intrinsic parameters.

[0060] By means of the above technical solution, the present invention provides a method for constructing a silicon steel quality prediction model based on a meta-heuristic optimization algorithm, which has at least the following beneficial effects:

[0061] 1. The present invention uses a BP neural network optimized based on a meta-heuristic algorithm to construct a quality prediction model to predict the quality of silicon steel, comprehensively considers multiple processes and multiple process parameters, realizes the prediction of product quality of all processes and multiple process parameters, and ensures the accuracy of model prediction.

[0062] 2. The present invention is applicable to a product quality prediction method in the silicon steel manufacturing process, which can realize accurate prediction of product quality. In subsequent production, key process parameters can be adjusted in time according to the quality prediction results, the manufacturing and assembly processes can be optimized, and the quality and consistency of products can be improved, thereby improving manufacturing efficiency and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0064] Figure 1 A flow chart of a method for constructing a silicon steel quality prediction model according to the present invention;

[0065] Figure 2 It is a data diagram of various parameters of silicon steel of the present invention. DETAILED DESCRIPTION

[0066] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific implementation methods, so that the implementation process of how the present application uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0067] Please refer to Figure 1 This embodiment proposes a method for constructing a silicon steel quality prediction model based on a meta-heuristic optimization algorithm. By using a BP neural network optimized based on a meta-heuristic algorithm, a quality prediction model is constructed to predict the quality of silicon steel. Multiple processes and multiple process parameters are comprehensively considered to achieve the prediction of product quality of all processes and multiple process parameters, thereby ensuring the accuracy of model prediction and solving the problems of difficulty in comprehensively predicting silicon steel quality and inaccuracy and low reliability of prediction results in the prior art. The method comprises the following steps:

[0068] S1, obtaining silicon steel data of overall silicon steel processing, and building a basic silicon steel prediction model based on the silicon steel data, the silicon steel data including the entire process and process parameters of the overall silicon steel processing; in step S1, the specific process includes the following steps:

[0069] S11, record the collected processing data of the entire process, and organize and mark them according to the corresponding relationship between the process and the process parameters;

[0070] S12, storing the collected processing data of the entire process into a database, and using the data storage method of the database to obtain complete silicon steel data;

[0071] S13. Use complete silicon steel data to train the BP neural network to obtain a basic silicon steel prediction model.

[0072] S2, determining the upper and lower limits of the process parameters according to any silicon steel production process standard, and setting the meta-heuristic optimization algorithm according to the upper and lower limits of the process parameters to initialize the relevant parameters of the particle population; in step S2, the specific process includes the following steps:

[0073] S21. Determine the upper and lower limits of the process parameters according to the production process standard of any silicon steel corresponding to the grade, and use the normalization method to normalize the upper and lower limits of the process parameters to between [-1,1];

[0074] S22, using formula The normalized upper and lower limits of the process parameters are used to generate random particle populations, where represents the position of the jth dimension of the nth population, rand represents a random number between (0,1), ub represents the upper bound of random number generation, i.e., upper bound; lb represents the lower bound, i.e., lower bound;

[0075] S23, initialize the relevant parameters x of the particle population n , x n The population matrix of the population in all dimensions can be depicted, namely:

[0076]

[0077] Where n = 1, 2, ...; N p =1,2,…,dim; j=1,2,…,dim.

[0078] S3, based on the FENR (Fitness-Enhanced Newton-Raphson) search rule, determine the position of the current individual particle in the particle population within the feasible region, and calculate the fitness value fitness according to the position of the current individual particle; in step S3, the specific process includes the following steps:

[0079] S31, using FENR search rules and related parameters x n+1 Allow particle populations to explore feasible areas more accurately and obtain better positions;

[0080]

[0081] x n+1 =x n -FENR

[0082] Among them, randn represents a normally distributed random number with a mean of 0 and a variance of 1; X w Indicates the worst position; X b Indicates the best position, Indicates the current location;

[0083] S32. Current location Update to get the current position of individual particles The update formula is:

[0084]

[0085] Among them, r 1 and r 2 are distinct integers randomly selected from the population; and They represent the position vectors of two different individuals randomly selected from the population; a and b are preset constants; fitness is the fitness value;

[0086] S33, according to the current position of individual particles Calculate the fitness value to select the optimal solution and further optimize the operation. The fitness value fitness calculation formula is:

[0087] fitness=-α×MSE+(1-α)×R 2

[0088]

[0089] Among them, y i is the actual measured value of the i-th sample, i.e., the target value; f(x i ) is the prediction value of the basic silicon steel prediction model for the i-th sample; MSE is the mean prediction error; R 2 is a fraction, indicating the proportion of variance explained by the basic silicon steel prediction model; α is a trade-off between the mean square error and R 2 The parameter of score contribution, the value range is (0≤α≤1); fitness is the fitness value; The sum of squares of the prediction errors of all samples is calculated; N is the total number of samples; is the average value of the actual measured values.

[0090] For each particle, compare its current fitness value fitness with the individual optimal fitness value. If the current fitness value fitness is better than the individual optimal fitness value, update the individual optimal position pfitness and the corresponding fitness value. At the same time, compare the current fitness value fitness with the global optimal fitness value bfitness. If it is better than the global optimal fitness value, update the global optimal position gfitness to adjust a better corresponding relationship. Where n is the number of features. Through multiple iterations, the correspondence between physical values ​​and features is gradually optimized through fitness.

[0091] S4, updating the global optimal position according to the fitness value fitness of the current individual particle to obtain the global optimal position, and inputting its position information into the basic silicon steel prediction model to realize the update of process parameters; since the current position of the individual particle is not the optimal position, it needs to be further optimized to obtain the global optimal position, so in step S4, the specific process includes the following steps:

[0092] S41, updating FENR search rules to update the global optimal position to obtain the global optimal position Specifically, the expression for updating the FENR search rule is:

[0093]

[0094] in,

[0095] y w =r 1 ×(Mean(Z n+1 +x n )+r 1 ×Δx)×fitness(Z n+1 )

[0096] y b =r 1 ×(Mean(Z n+1 +x n )-r 1 ×Δx)×fitness(Z n+1 )

[0097]

[0098] In the above formula, Z n+1 is the next position calculated based on the current position and the random number; y w and b Use Z n+1 and x n The positions of the two vectors generated; (fitness(y w )-fitness(y b )) is the fitness difference, which is used to weight the FENR search rule, emphasizing that the positions with higher fitness contribute more to the update; r 1 Represents a random number between (0,1); x n are related parameters; X w Indicates the worst position; X b Indicates the best position, Indicates the current location.

[0099] Global optimal position The expression is:

[0100]

[0101] in, and They represent the position vectors of two different individuals randomly selected from the population; a and b are preset constants; To update the current location Get the current position of the individual particle.

[0102] S42, the global optimal position Bring it into the trap avoidance algorithm based on fitness value to get the location information Location Information The calculation formula is:

[0103]

[0104] μ 1 =β×3×rand+(1-β)

[0105] μ 2 =β×rand+(1-β)

[0106] Among them, θ 1 ,θ 2 are uniform random numbers between (-1,1) and (-0.5,0.5); μ 1 , μ 2 is a random number; rand is a uniform random number between (0,1); β is the exploration adjustment factor, and its value range is [0,1];

[0107] S43, the global optimal position Location information It is used as input to the basic silicon steel prediction model to update process parameters.

[0108] In this embodiment, the Fitness-Enhanced Newton-Raphson meta-heuristic optimization algorithm is used to optimize the global optimal position and obtain the current individual optimal position. Location information Then it is used as the input of the BP neural network, thereby improving the precision and accuracy of feature extraction by the BP neural network.

[0109] S5. Determine whether the global optimal position is within the process standard range according to the upper and lower limits of the process parameters; if so, end; the global optimal position represents the optimal process parameters. At this time, any corresponding optimal process parameter obtained must meet the process standards, that is, it cannot exceed the upper and lower limits of the process parameters. Therefore, the judgment here only needs to determine that the global optimal position is within the upper and lower limits of the process parameters to meet the requirements. The upper and lower limits of the process parameters need to be determined according to the specific actual process standards, or directly obtained and determined from the corresponding process standards. This embodiment does not make any restrictions here, and the obtained basic silicon steel prediction model can realize the prediction of silicon steel quality for all processes and multiple process parameters.

[0110] If not, the internal parameters of the basic silicon steel prediction model are optimized, and then the process returns to step S3; in this embodiment, the process parameters to be optimized are selected according to the characteristics of silicon steel steel loss, magnetic field strength (B) and magnetization capacity. The process parameters can be specific physical or chemical parameters, such as processing temperature, processing time, alloy composition, etc. Among them, under the magnetic field strength (B), the magnetization capacity corresponding to the unit iron loss of 23ZDKH90 in the silicon steel steel loss, the magnetization capacity corresponding to the unit iron loss of 23ZH90, the magnetization capacity corresponding to the unit iron loss of Z7H-0.27, and the magnetization capacity corresponding to the unit iron loss of 27ZH100, such as Figure 2 shown.

[0111] In step S5, the internal parameters of the basic silicon steel prediction model are optimized, and the specific process includes the following steps:

[0112] S51, extract the internal parameters of the basic silicon steel prediction model and save them as parameter set X and encode them to obtain binary-coded internal parameters; here, the internal parameters need to be preprocessed first, including data cleaning, missing value processing, standardization or normalization, etc., to ensure the quality and consistency of the data. Then encode the selected internal parameters appropriately so that the subsequent meta-heuristic algorithm can process them. If you choose to use real number encoding, use the numerical value of the internal parameter directly; if you use binary encoding, you need to convert the internal parameter into binary representation, and binary encoding is used in this embodiment.

[0113] S52, set the number of neurons n in the input layer of the BP neural network input =3, the number of neurons in the hidden layer n hidden =10, the number of neurons in the output layer n output =1, the weight matrix between the input layer and the hidden layer is W1, the weight matrix between the hidden layer and the output layer is W2, and the bias vectors of the hidden layer and the output layer are b1 and b2 respectively;

[0114] S53, decoding the binary-coded internal parameters into real values, and mapping them to obtain optimized process internal parameters, which includes mapping the decoded process parameters to a real value of size n. hidden ×n input Each element position of the weight matrix W1 is mapped to a size of n output ×n hidden Each element position of the weight matrix W2 and the length of the mapping are n hidden and n output Each element position of the bias vectors b1 and b2 is finally obtained to obtain the optimized intrinsic parameters.

[0115] The present invention is applicable to a product quality prediction method in a silicon steel manufacturing process, and can realize accurate prediction of product quality. In subsequent production, key process parameters can be adjusted in a timely manner according to the quality prediction results, the manufacturing and assembly processes can be optimized, and the quality and consistency of products can be improved, thereby improving manufacturing efficiency and economic benefits.

[0116] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, so the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0117] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the above embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0118] The above implementation methods have been described in detail. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for constructing a silicon steel quality prediction model based on a meta-heuristic optimization algorithm, characterized in that: The method comprises the following steps: S1. Obtain silicon steel data of overall silicon steel processing, and build a basic silicon steel prediction model based on the silicon steel data. The silicon steel data includes all the processes and process parameters of overall silicon steel processing. The specific process includes the following steps: S11, record the collected processing data of the entire process, and organize and mark them according to the corresponding relationship between the process and the process parameters; S12, storing the collected processing data of the entire process into a database, and using the data storage method of the database to obtain complete silicon steel data; S13, using complete silicon steel data to train the BP neural network to obtain a basic silicon steel prediction model; S2. Determine the upper and lower limits of process parameters according to any silicon steel production process standard, and set the relevant parameters of the meta-heuristic optimization algorithm for initializing the particle population according to the upper and lower limits of the process parameters; S3, determine the position of the current individual particle in the particle population within the feasible area based on the FENR search rule, and calculate the fitness value fitness according to the position of the current individual particle; S4, updating the global optimal position according to the fitness value fitness of the current individual particle to obtain the global optimal position, and inputting its position information into the basic silicon steel prediction model to realize the update of process parameters; S5. judging whether the global optimal position is within the process standard range according to the upper and lower limits of the process parameters; If so, then end; If not, the parameters of the basic silicon steel prediction model are optimized, and then the process returns to step S3; The internal parameters of the basic silicon steel prediction model are optimized. The specific process includes the following steps: S51, extracting the internal parameters of the basic silicon steel prediction model and saving them as a parameter set X and encoding them to obtain binary-coded internal parameters; S52, set the number of neurons n in the input layer of the BP neural network input =3, the number of neurons in the hidden layer n hidden =10, the number of neurons in the output layer n output =1, the weight matrix between the input layer and the hidden layer is W1, the weight matrix between the hidden layer and the output layer is W2, and the bias vectors of the hidden layer and the output layer are b1 and b2 respectively; S53, decoding the binary-coded internal parameters into real values, and mapping them to obtain optimized process internal parameters, which includes mapping the decoded process parameters to a real value of size n. hidden ×n input Each element position of the weight matrix W1 is mapped to a size of n output ×n hidden Each element position of the weight matrix W2 and the length of the mapping are n hidden and n output Each element position of the bias vectors b1 and b2 is finally obtained to obtain the optimized intrinsic parameters.

2. The method for constructing a silicon steel quality prediction model according to claim 1, characterized in that: In step S2, the specific process includes the following steps: S21. Determine the upper and lower limits of the process parameters according to the production process standard of any silicon steel corresponding to the grade, and use the normalization method to normalize the upper and lower limits of the process parameters to between [-1,1]; S22, using formula The normalized upper and lower limits of the process parameters are used to generate random particle populations, where represents the position of the jth dimension of the nth population, rand represents a random number between (0,1), ub represents the upper bound of random number generation, i.e., upper bound; lb represents the lower bound, i.e., lower bound; S23, initialize the relevant parameters x of the particle population n , x n The population matrix of the population in all dimensions can be depicted, namely: Where n = 1, 2, ...; N p =1,2,…,dim; j=1,2,…,dim.

3. The method for constructing a silicon steel quality prediction model according to claim 1, characterized in that: In step S3, the specific process includes the following steps: S31, using FENR search rules and related parameters x n+1 Allow particle populations to explore feasible areas more accurately and obtain better positions; x n+1 =x n -FENR Among them, randn represents a normally distributed random number with a mean of 0 and a variance of 1; X w Indicates the worst position; X b Indicates the best position, Indicates the current location; S32. Current location Update to get the current position of individual particles The update formula is: Where r1 and r2 are different integers randomly selected from the population; and They represent the position vectors of two different individuals randomly selected from the population; a and b are preset constants; fitness is the fitness value; S33, according to the current position of individual particles Calculate the fitness value fitness in order to select the best solution and use it for further optimization operations.

4. The method for constructing a silicon steel quality prediction model according to claim 3, characterized in that: In step S33, the calculation formula of the fitness value fitness is: fitness=-α×MSE+(1-α)×R 2 Among them, y i is the actual measured value of the i-th sample, i.e., the target value; f(x i ) is the prediction value of the basic silicon steel prediction model for the i-th sample; MSE is the mean prediction error; R 2 is a fraction, indicating the proportion of variance explained by the basic silicon steel prediction model; α is a trade-off between the mean square error and R 2 The parameter of the score contribution has a value range of (0≤α≤1); fitness is the fitness value; N is the total number of samples; is the average value of the actual measured values.

5. The method for constructing a silicon steel quality prediction model according to claim 1, characterized in that: In step S4, the specific process includes the following steps: S41, updating FENR search rules to update the global optimal position to obtain the global optimal position Specifically, the expression for updating the FENR search rule is: in, y w =r1×(Mean(Z n+1 +x n )+r1×Δx)×fitness(Z n+1 ) y b =r1×(Mean(Z n+1 +x n )-r1×Δx)×fitness(Z n+1 ) In the above formula, Z n+1 is the next position calculated based on the current position and the random number; y w and b Use Z n+1 and x n The positions of the two vectors generated; (fitness(y w )-fitness(y b )) is the fitness difference, which is used to weight the FENR search rule, emphasizing that the position with higher fitness contributes more to the update; r1 represents a random number between (0,1); x n are related parameters; X w Indicates the worst position; X b Indicates the best position, Indicates the current position; randn indicates a normally distributed random number with a mean of 0 and a variance of 1; Global optimal position The expression is: in, and They represent the position vectors of two different individuals randomly selected from the population; a and b are preset constants; To update the current location The position of the current individual particle obtained; S42, the global optimal position Bring it into the trap avoidance algorithm based on fitness value to get the location information S43, the global optimal position Location information It is used as input to the basic silicon steel prediction model to update process parameters.

6. The method for constructing a silicon steel quality prediction model according to claim 5, characterized in that: In step S42, the location information The calculation formula is: μ1=β×3×rand+(1-β) μ2=β×rand+(1-β) Among them, θ1 and θ2 are consistent random numbers between (-1,1) and (-0.5,0.5) respectively; μ1 and μ2 are random numbers; rand is a uniform random number between (0,1); β is the exploration adjustment factor, and its value range is [0,1].

Images