Method for optimizing chemical components of plate

Abstract

The invention discloses a plate chemical component optimization method which comprises the following steps: S1, utilizing multiple linear regression analysis to obtain chemical elements which have small influence on yield strength and tensile strength of steel; s2, constructing a BP neural network model by taking the chemical elements with small influence as input variables and taking the yield strength and tensile strength of the steel as output variables; s3, the chemical elements with small influences serve as input variables, the steel smelting cost serves as an output variable, and a smelting cost mathematical model is constructed; s4, the smelting cost mathematical model is utilized, a genetic algorithm is constructed, and automatic optimization of the chemical components of all the elements is achieved with the purpose of smelting cost minimization. According to the method provided by the invention, the chemical components are optimized based on the existing rolling data, additional smelting is not needed, the optimization cost is relatively low, the quantity of samples capable of being processed is large, the result is more accurate and reliable, and the optimization time is greatly shortened.

Description

Technical Field

[0001] The invention relates to the technical field of metallurgy, and in particular to a method for optimizing the chemical composition of a plate. Background Art

[0002] As an important steel product, hot-rolled plate has formed a huge market scale in China. Its downstream fields are extensive, mainly concentrated in cold-rolled coils, machinery, building steel structures, containers, steel pipes, automobiles, home appliances and other fields. In order to increase market share, various steel companies are optimizing the rolling and smelting process to reduce production costs.

[0003] Patent No.: CN201210193364.7 discloses a production method of x70 hot-rolled medium and thick plates for submarine pipelines. This method mainly optimizes the rolling process and does not study the chemical composition. Patent No.: CN 200510023770 discloses a method for improving the prediction accuracy of hot rolling force using strip chemical composition data. This method mainly predicts the rolling force based on the chemical composition of the strip, thereby improving the stability of rolling and the accuracy of thickness control. Patent No.: CN201810147727.0 discloses a 540MPa grade hot-rolled pickled steel plate and its manufacturing method. This method is mainly based on a limited number of industrial experiments to obtain the chemical composition of 540MPa hot-rolled steel plates. As we all know, industrial experiments are not only inefficient, but also costly. In addition, the optimization results may not be accurate only through a limited number of experiments.

[0004] In summary, the existing methods cannot effectively optimize the chemical composition of the board. Therefore, it is urgent to develop a chemical composition optimization method suitable for the board. Summary of the invention

[0005] The purpose of the present invention is to provide a method for optimizing the chemical composition of a plate in order to overcome the defects of the above-mentioned prior art.

[0006] The present invention solves the technical problem by adopting the following technical solutions.

[0007] The present invention provides a method for optimizing the chemical composition of a plate, comprising the following steps:

[0008] S1. Using multiple linear regression analysis, we can find the chemical elements that have the least influence on the yield and tensile strength of steel.

[0009] S2, taking the chemical elements with less influence as input variables, and the yield strength and tensile strength of steel as output variables, constructing a BP neural network model;

[0010] S3. Take the chemical elements with less influence as input variables and the steel smelting cost as output variable to construct a mathematical model of smelting cost;

[0011] S4. Utilize the mathematical model of smelting cost and construct a genetic algorithm to automatically optimize the chemical composition of each element with the goal of minimizing the smelting cost.

[0012] The present invention has the following beneficial effects:

[0013] The present invention provides a plate chemical composition optimization method. The chemical composition optimization is performed based on existing rolling data, and no additional smelting is required, so the optimization cost is low; compared with limited experiments, the accuracy and reliability of the optimization results are greatly improved; from data collation to optimization result output, only 1 to 2 hours are required, and compared with industrial experiments, the optimization efficiency is significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 Flowchart of the method for optimizing the chemical composition of the board;

[0016] Figure 2 Schematic diagram of genetic algorithm optimization. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the technical scheme in the embodiments of the present invention will be described clearly and completely below. If the specific conditions are not specified in the embodiments, they are carried out according to conventional conditions or conditions recommended by the manufacturer. If the manufacturer of the reagents or instruments used is not specified, they are all conventional products that can be purchased commercially.

[0018] A method for optimizing the chemical composition of a plate material provided by an embodiment of the present invention is described in detail below.

[0019] In a first aspect, an embodiment of the present invention provides a method for optimizing the chemical composition of a plate, comprising the following steps:

[0020] S1. Using multiple linear regression analysis, we can find the chemical elements that have the least influence on the yield and tensile strength of steel.

[0021] S2, taking the chemical elements with less influence as input variables, and the yield strength and tensile strength of steel as output variables, constructing a BP neural network model;

[0022] S3. Take the chemical elements with less influence as input variables and the steel smelting cost as output variable to construct a mathematical model of smelting cost;

[0023] S4. Utilize the mathematical model of smelting cost and construct a genetic algorithm to automatically optimize the chemical composition of each element with the goal of minimizing the smelting cost.

[0024] Wherein, the step S1 comprises the following steps:

[0025] S11. Collect target data during the sheet production process;

[0026] S12. Sort the target data, delete unnecessary parts, and retain only the main element composition and the corresponding yield strength and tensile strength.

[0027] S13, taking the above-mentioned element composition as the independent variable and the plate yield strength as the dependent variable, a multi-factor regression analysis is performed to obtain the chemical elements with the least influence on the yield strength;

[0028] S14, taking the above-mentioned element composition as the independent variable and the tensile strength of the plate as the dependent variable, a multi-factor regression analysis is performed again to obtain the chemical elements with the least influence on the tensile strength;

[0029] S15. Combining the results of S13 and S14, select elements that have a smaller impact on yield strength and tensile strength, and use the above elements as input variables of the BP neural network algorithm.

[0030] The step S2 comprises the following steps:

[0031] The chemical elements with less influence are taken as input variables, and the yield and tensile strength of steel are taken as output variables to construct a BP neural network model, in which: the number of input layer nodes is 5, the number of output layer nodes is 2, the number of neural network layers is 3, and the number of hidden layer nodes is 11.

[0032] The step S4 comprises the following steps:

[0033] S41, setting the genetic algorithm population size, the number of iterations, and the range of each element content;

[0034] S42, establishing an adaptive function, i.e., a mathematical model of smelting cost;

[0035] S43, setting population boundary conditions, that is, after each population is generated, it is first substituted into the BP neural network. If the yield strength or tensile strength is less than the value specified by the national standard, the population is discarded. When the yield strength and tensile strength predicted values ​​are greater than the values ​​specified by the national standard, the population is brought into the adaptive function to obtain the adaptive value;

[0036] S44, automatically searching for the best solution to obtain the final composition of each element.

[0037] The present invention will be further described below in conjunction with the embodiments.

[0038] Example 1

[0039] Take ship plate EH40 as an example to optimize the chemical composition of hot rolled plate, see Figure 1 , including the following steps:

[0040] S1. Use multiple linear regression analysis to obtain the chemical elements that have the least influence on the yield strength and tensile strength of steel; the specific steps are:

[0041] S11. Use the backend software to export 55,000 copies of the one-year production data of EH40 ship plates.

[0042] S12. Sort the data, delete unnecessary parts, and only keep the elemental compositions such as C, Si, Mn, Al, Cr, Cu, Ti, V, Mo, Nb, Ni, Ca and the corresponding yield and tensile strengths.

[0043] Table 1 Sensitivity coefficients of chemical elements when yield strength is the dependent variable

[0044]

[0045] S13. Taking the above-mentioned elemental composition as the independent variable and the plate yield strength as the dependent variable, a multi-factor regression analysis is performed to obtain the chemical elements with the least influence on the yield strength, which are Si, Cr, Ni, Ti, Mn, Mo, Al, etc.;

[0046] Table 2 Sensitivity coefficients of chemical elements when tensile strength is the dependent variable

[0047]

[0048] S14. Taking the tensile strength of the plate as the dependent variable, multi-factor regression analysis was performed again, and the chemical elements with the least influence on the tensile strength were obtained, namely Si, Ni, Cr, Ti, Ca, Mo, V, etc.;

[0049] S15. Combining the results of S13 and S14, select elements that have less influence on yield strength and tensile strength, namely Si, Ni, Cr, Ti, and Mo, and use the above elements as input variables of the neural network algorithm.

[0050] S2. The chemical elements with less influence are used as input variables, and the yield strength and tensile strength of steel are used as output variables to construct a BP neural network model; among which: the number of input layer nodes is 5, the number of output layer nodes is 2, the number of neural network layers is 3, the number of hidden layer nodes is 11, the activation function is the S-type tangent function Tansig, the learning efficiency is 0.01, and the target error is 1×10 -6 , the number of iterations is 1000. The number of training samples is 50,000 and the number of test samples is 5000.

[0051] S3. Construct a mathematical model of chemical elements and smelting costs, as follows:

[0052] N=100×CSi+1300×CNi+1600×CCr+6000×CTi+560×CMo

[0053] Where N is the smelting cost, C Si , C Ni , C Cr , C Ti , C Mo Chemical composition (%): Si, Ni, Cr, Ti, Mo.

[0054] S4. Using the mathematical model of smelting cost, a genetic algorithm is constructed to minimize the smelting cost and realize the automatic optimization of the chemical composition of each element. Figure 2 , the specific steps are:

[0055] S41. Set the population size of the genetic algorithm to 20, the number of iterations to 2000, and the range of the content of each element to [0, 2%].

[0056] S42 establishes an adaptive function, namely, a mathematical model of smelting cost.

[0057] S43 sets the population boundary conditions, that is, after each population is generated, it is first substituted into the BP neural network. If the yield or tensile strength is less than the national standard value, that is, the adaptive value is 0, the population is discarded. When the yield and tensile strength prediction values ​​are greater than the national standard values ​​of 390MPa and 510MPa, the population is then brought into the adaptive function to calculate the adaptive value.

[0058] S44 automatically optimizes and obtains the final composition of each element, which is C Si =0.18%, C Ni =0.22%, C Cr =0.02%, C Ti =0.014%, C Mo =0.03%, at this time, the optimized smelting cost is 436.8 yuan / ton.

[0059] Before optimization, the chemical composition of EH40 ship plate is: Si =0.34%, C Ni =0.32%, C Cr =0.12%, C Ti =0.018%, C Mo =0.05%, the smelting cost corresponding to this component is 778 yuan.

[0060] Comprehensive comparison shows that the smelting cost of EH40 ship plate optimized by the above method can be reduced by 341.2 yuan while ensuring the yield and tensile strength. According to the annual production capacity of 200,000 tons of steel enterprises, additional economic benefits of 68.24 million yuan are generated.

[0061] Example 2

[0062] Taking ship plate Q420 as an example, the optimization of chemical composition of hot-rolled plate is explained in detail.

[0063] S1. Use multiple linear regression analysis to obtain the chemical elements that have the least influence on the yield and tensile strength of steel; the specific steps are:

[0064] S11. Use the backend software to export 55,000 copies of the one-year production data of Q420 ship plates.

[0065] S12. Sort the data, delete unnecessary parts, and only keep the elemental compositions such as C, Si, Mn, Al, Cr, Cu, Ti, V, Mo, Nb, Ni, Ca and the corresponding yield and tensile strengths.

[0066] S13. Taking the above-mentioned elemental composition as the independent variable and the plate yield strength as the dependent variable, a multi-factor regression analysis is performed to obtain the chemical elements with the least influence on the yield strength, which are C, Si, Cr, Ni, Ti, Mn, Mo, Ca, etc.;

[0067] S14. Taking the tensile strength of the plate as the dependent variable, multi-factor regression analysis was performed again, and the chemical elements with the smallest influence on the tensile strength were obtained as follows: Si, Ni, Cr, Ti, Ca, Mo, V, etc.;

[0068] S15. Combining the results of S13 and S14, select elements with less influence on yield strength and tensile strength, namely Si, Ni, Cr, Ti, and Mo. The above elements are used as inputs of the neural network algorithm.

[0069] S2. The chemical elements with less influence are used as input variables, and the yield strength and tensile strength of steel are used as output variables to construct a BP neural network model. The number of input layer nodes is 5, the number of output layer nodes is 2, the number of neural network layers is 4, the number of hidden layer nodes is 11, the activation function is the S-type tangent function Tansig, the learning efficiency is 0.01, and the target error is 1×10 -6 , the number of iterations is 1000. The number of training samples is 50,000 and the number of test samples is 5000.

[0070] S3. Construct a mathematical model of chemical elements and smelting costs, as follows:

[0071] N=100×CSi+1300×CNi+1600×CCr+6000×CTi+560×CMo

[0072] Where N is the smelting cost, C Si , C Ni , C Cr , C Ti , C Mo Chemical composition (%): Si, Ni, Cr, Ti, Mo.

[0073] S4. Using the mathematical model of smelting cost, a genetic algorithm is constructed to minimize the smelting cost and automatically optimize the chemical composition of each element. The specific steps are:

[0074] S41. Set the population size of the genetic algorithm to 25, the number of iterations to 2500, and the range of the content of each element to [0, 2%].

[0075] S42 establishes an adaptive function, namely, a mathematical model of smelting cost.

[0076] S43 sets the population boundary conditions, that is, after each population is generated, it is first substituted into the BP neural network. If the yield or tensile strength is less than the national standard value, that is, the adaptive value is 0, the population is discarded. When the yield and tensile strength prediction values ​​are greater than the national standard values ​​of 390MPa and 510MPa, the population is then brought into the adaptive function to calculate the adaptive value.

[0077] S44 automatically optimizes and obtains the final composition of each element, which is C Si =0.23%, C Ni =0.22%, C Cr =0.03%, C Ti =0.012%, C Mo =0.04%, at this time, the optimized smelting cost is 451.4 yuan / ton.

[0078] Before optimization, the chemical composition of EH40 ship plate is: Si=0.36%, C Ni =0.28%, C Cr =0.14%, C Ti =0.02%, C Mo =0.06%, the smelting cost corresponding to this component is 777.6 yuan.

[0079] Comprehensive comparison shows that the smelting cost of Q420 plate optimized by the above method can be reduced by 326.2 yuan while ensuring the yield and tensile strength. According to the annual production capacity of 100,000 tons of steel enterprises, additional economic benefits of 32.62 million yuan are generated.

[0080] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for optimizing the chemical composition of a plate, characterized in that: The following steps are involved: S1. Using multiple linear regression analysis, we can find the chemical elements that have the least influence on the yield and tensile strength of steel. S2. Using the chemical elements with less influence as input variables and the yield strength and tensile strength of steel as output variables, a BP neural network model is constructed; S3, taking the chemical elements with less influence as input variables and the steel smelting cost as output variables, to construct a smelting cost mathematical model; S4. Using the mathematical model of smelting cost and constructing a genetic algorithm, the automatic optimization of the chemical composition of each element is achieved with the goal of minimizing the smelting cost.

2. The method for optimizing the chemical composition of a plate according to claim 1, characterized in that: The step S1 comprises the following steps: S11. Collect target data during the sheet production process; S12. Arrange the target data, delete unnecessary parts, and retain only the main element composition and the corresponding yield strength and tensile strength; S13, taking the above-mentioned element composition as the independent variable and the plate yield strength as the dependent variable, a multi-factor regression analysis is performed to obtain the chemical elements with the least influence on the yield strength; S14, taking the above-mentioned element composition as the independent variable and the tensile strength of the plate as the dependent variable, a multi-factor regression analysis is performed again to obtain the chemical elements with the least influence on the tensile strength; S15. Combining the results of S13 and S14, select elements that have a smaller impact on yield strength and tensile strength, and use the above elements as input variables of the BP neural network algorithm.

3. The method for optimizing the chemical composition of a plate according to claim 1, characterized in that: The step S2 comprises the following steps: The chemical elements with less influence are taken as input variables, and the yield and tensile strength of steel are taken as output variables to construct the BP neural network model.

4. The method for optimizing the chemical composition of a plate material according to claim 3, characterized in that: The number of input layer nodes is 5, the number of output layer nodes is 2, the number of neural network layers is 3, and the number of hidden layer nodes is 11.

5. The method for optimizing the chemical composition of a plate according to claim 1, characterized in that: The step S4 comprises the following steps: S41, setting the genetic algorithm population size, the number of iterations, and the range of each element content; S42, establishing an adaptive function, i.e., a mathematical model of smelting cost; S43, setting population boundary conditions, that is, after each population is generated, it is first substituted into the BP neural network. If the yield strength or tensile strength is less than the value specified by the national standard, the population is discarded. When the yield strength and tensile strength predicted values ​​are greater than the values ​​specified by the national standard, the population is brought into the adaptive function to obtain the adaptive value; S44, automatically searching for the best solution to obtain the final composition of each element.

6. The method for optimizing the chemical composition of a plate material according to claim 5, characterized in that: The population size of the genetic algorithm is 20, the number of iterations is 2000, and the content of each element ranges from [0, 2%].

Images