CN117877617A - Method for determining addition amount of converter alloy by combining differential evolution and whale algorithm - Google Patents
Method for determining addition amount of converter alloy by combining differential evolution and whale algorithm Download PDFInfo
- Publication number
- CN117877617A CN117877617A CN202410050745.2A CN202410050745A CN117877617A CN 117877617 A CN117877617 A CN 117877617A CN 202410050745 A CN202410050745 A CN 202410050745A CN 117877617 A CN117877617 A CN 117877617A
- Authority
- CN
- China
- Prior art keywords
- neural network
- converter
- alloy
- woa
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 229910045601 alloy Inorganic materials 0.000 title claims abstract description 46
- 239000000956 alloy Substances 0.000 title claims abstract description 46
- 238000000034 method Methods 0.000 title claims abstract description 46
- 241000283153 Cetacea Species 0.000 title claims abstract description 26
- 230000008569 process Effects 0.000 claims abstract description 25
- 238000013528 artificial neural network Methods 0.000 claims abstract description 23
- 229910000831 Steel Inorganic materials 0.000 claims abstract description 18
- 239000010959 steel Substances 0.000 claims abstract description 18
- 238000004519 manufacturing process Methods 0.000 claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 15
- 238000010079 rubber tapping Methods 0.000 claims abstract description 10
- 238000009628 steelmaking Methods 0.000 claims abstract description 10
- 238000005457 optimization Methods 0.000 claims abstract description 8
- 238000007792 addition Methods 0.000 claims abstract 13
- 238000003062 neural network model Methods 0.000 claims description 43
- 230000006870 function Effects 0.000 claims description 22
- 210000002569 neuron Anatomy 0.000 claims description 17
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 claims description 14
- 230000007246 mechanism Effects 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 9
- 238000003723 Smelting Methods 0.000 claims description 8
- 229910052742 iron Inorganic materials 0.000 claims description 7
- 229910052799 carbon Inorganic materials 0.000 claims description 5
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical group [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 claims description 4
- DALUDRGQOYMVLD-UHFFFAOYSA-N iron manganese Chemical compound [Mn].[Fe] DALUDRGQOYMVLD-UHFFFAOYSA-N 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 4
- MBMLMWLHJBBADN-UHFFFAOYSA-N Ferrous sulfide Chemical compound [Fe]=S MBMLMWLHJBBADN-UHFFFAOYSA-N 0.000 claims description 3
- 102000005298 Iron-Sulfur Proteins Human genes 0.000 claims description 3
- 108010081409 Iron-Sulfur Proteins Proteins 0.000 claims description 3
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical group [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 claims description 3
- 229910000805 Pig iron Inorganic materials 0.000 claims description 3
- 230000003247 decreasing effect Effects 0.000 claims description 3
- XWHPIFXRKKHEKR-UHFFFAOYSA-N iron silicon Chemical compound [Si].[Fe] XWHPIFXRKKHEKR-UHFFFAOYSA-N 0.000 claims description 3
- QMQXDJATSGGYDR-UHFFFAOYSA-N methylidyneiron Chemical compound [C].[Fe] QMQXDJATSGGYDR-UHFFFAOYSA-N 0.000 claims description 3
- 230000035772 mutation Effects 0.000 claims description 3
- DPTATFGPDCLUTF-UHFFFAOYSA-N phosphanylidyneiron Chemical compound [Fe]#P DPTATFGPDCLUTF-UHFFFAOYSA-N 0.000 claims description 3
- 229910052698 phosphorus Inorganic materials 0.000 claims description 3
- 239000011574 phosphorus Substances 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 claims description 2
- 238000012937 correction Methods 0.000 claims description 2
- 230000008901 benefit Effects 0.000 abstract description 8
- 238000009851 ferrous metallurgy Methods 0.000 abstract description 2
- 230000035945 sensitivity Effects 0.000 abstract description 2
- 239000002245 particle Substances 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- PXHVJJICTQNCMI-UHFFFAOYSA-N Nickel Chemical compound [Ni] PXHVJJICTQNCMI-UHFFFAOYSA-N 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000005275 alloying Methods 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000002028 premature Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000000047 product Substances 0.000 description 2
- 230000011273 social behavior Effects 0.000 description 2
- 229910000851 Alloy steel Inorganic materials 0.000 description 1
- VYZAMTAEIAYCRO-UHFFFAOYSA-N Chromium Chemical compound [Cr] VYZAMTAEIAYCRO-UHFFFAOYSA-N 0.000 description 1
- 229910000616 Ferromanganese Inorganic materials 0.000 description 1
- 229910000519 Ferrosilicon Inorganic materials 0.000 description 1
- 238000010220 Pearson correlation analysis Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 229910052804 chromium Inorganic materials 0.000 description 1
- 239000011651 chromium Substances 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 229910052759 nickel Inorganic materials 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Abstract
The invention discloses a method for determining the addition of a converter alloy by combining differential evolution and whale algorithm, and belongs to the technical field of ferrous metallurgy. The invention comprises the following steps: collecting a converter production dataset; fusing differential evolution and whale algorithm to obtain DE-WOA optimization algorithm; optimizing the BP neural network by using a DE-WOA algorithm to obtain the DE-WOA-BP neural network; training by using a DE-WOA-BP neural network to obtain a model; alloy additions were calculated using model predictions. The invention mainly solves the technical problems of low alloy addition control precision, low steel tapping molten steel component control precision and high alloy input cost in the converter steelmaking tapping process caused by the sensitivity to initial weight and learning rate when the existing BP neural network is used for determining the converter alloy addition, and improves the molten steel quality and economic benefit.
Description
Technical Field
The invention belongs to the technical field of ferrous metallurgy, and particularly relates to a method for determining the addition amount of a converter alloy by combining differential evolution and whale algorithm.
Background
Alloy steel is an alloy composed of iron and other elements (e.g., carbon, chromium, nickel, etc.). By adding different alloy elements into the common steel, the physical and chemical properties of the common steel can be changed, so that the strength, hardness, wear resistance, corrosion resistance and other characteristics of the common steel are improved. Alloy steels generally have high strength, good toughness and plasticity, making them widely used in the industry for manufacturing various parts and components, especially parts that are required to withstand high loads and severe operating environments.
In the conventional converter steelmaking process, determining the alloy addition amount mainly depends on manual experience. The operator selects the appropriate alloy type according to the process requirements for smelting the steel grade and approximately estimates the required alloy weight. However, the method has the problems of low accuracy and poor economic benefit, and the problems of repeated supplement or alloy waste in the later stage of steelmaking caused by inaccurate alloy addition amount estimation often occur, and even the situation of exceeding of components can be caused in severe cases. Therefore, there is a need to improve the determination of the amount of alloy added to improve accuracy and economic efficiency.
At present, the method for predicting the alloy addition in the steel tapping process of converter steelmaking mainly comprises a particle swarm algorithm and a BP neural network. The particle swarm algorithm (Particle Swarm Optimization, PSO) is a heuristic optimization algorithm, and inspiration is derived from the behaviors of social groups such as shoals or shoals. The particle swarm algorithm has the advantages of simplicity, easiness in implementation and high calculation efficiency, is very sensitive to an initial solution of a problem, can be trapped in a local optimal solution and cannot reach a global optimal solution, and can be limited in searching capacity for a complex high-dimensional optimization problem; the BP neural network (Back Propagation Neural Network, BPNN) is a common feedforward artificial neural network model widely applied to the fields of machine learning and neural networks, can model and approximate various nonlinear functions, and has strong generalization capability. However, BP neural networks also have some drawbacks, such as being prone to be trapped in locally optimal solutions, being more sensitive to initial weights and learning rates, etc.
Disclosure of Invention
1. Technical problem to be solved by the invention
The invention mainly aims to provide a method for determining the addition of a converter alloy by combining differential evolution and whale algorithm, and aims to solve the problems of low steelmaking accuracy and poor economic benefit caused by sensitivity to initial weight and learning rate when a BP neural network is used for determining the addition of the converter alloy.
2. Technical proposal
In order to achieve the above purpose, the technical scheme provided by the invention is as follows:
the invention discloses a method for determining the addition of a converter alloy by fusing differential evolution and whale algorithm, which comprises the following steps:
s1, collecting converter production environment data, and constructing a database for training a neural network model and storing model prediction results after manual correction;
s2, according to a metallurgical principle, establishing a process variable which influences the alloy addition amount in the tapping process of the converter in a database and taking the process variable as the input amount of a neural network model;
s3, preprocessing the data, creating a BP neural network model of a three-layer network layer, and using the obtained preprocessed converter production environment data set for training the neural network model;
s4, iterating the neuron number, the learning rate, the weight and the bias parameters of the BP neural network model established in the step S3 by using a DE-WOA algorithm and an fitness function, evaluating the performance of the parameter combination according to the fitness function return value, and applying the parameter combination with the minimum return value as the optimal parameter combination to the BP neural network to obtain the DE-WOA-BP neural network model;
s5, training and testing the DE-WOA-BP neural network model;
s6, collecting real-time process variables of the required converter smelting process, predicting the process variables by using a DE-WOA-BP neural network model, and calculating the type and weight of the alloy to be added in the converter smelting process; and (3) carrying out alloy addition according to the prediction result, and after tapping, storing the data into a database of the DE-WOA-BP neural network model.
Further, in the step S2, the input amount of the neural network model after analysis according to the correlation between the converter steelmaking reaction mechanism and Pearson specifically includes:
molten iron carbon content, molten iron silicon content, molten iron manganese content, molten iron phosphorus content, molten iron sulfur content, molten iron temperature, molten iron loading, scrap steel loading, pig iron loading, terminal temperature, terminal carbon content, terminal phosphorus content and alloy weight.
Still further, the data preprocessing operation in step S3 includes data cleansing and data reduction.
Further, the BP neural network model newly created in step S3 includes an input layer, an hidden layer and an output layer, wherein the input layer uses a fully connected layer having 12 neurons and uses a ReLU as an activation function; the hidden layer uses a fully connected layer with 6 neurons and uses ReLU as an activation function; the output layer uses a fully connected layer with 1 neuron and uses a linear activation function.
Further, the specific steps for obtaining the DE-WOA-BP neural network model in the step S4 are as follows:
t1, setting the size of a whale population, randomly initializing the population, calculating individual fitness f and average fitness aver of the population, optimizing the objective to minimize error, and setting an individual fitness function as the inverse of root mean square error;
t2, if f is less than or equal to aver, introducing a differential evolution algorithm to optimize a whale algorithm, sequentially performing mutation, crossover and selection operations, and applying a global search strategy to avoid trapping in local optimization;
if f > aver, replacing the current individual as the optimal individual in the population, and calculating a coefficient A and a random number p;
t4 if p is more than or equal to 0.5, applying a spiral update mechanism;
t5. if p <0.5 and |a| <1, applying a shrink wrap mechanism;
t6. if p <0.5 and |a| is not less than 1, applying a random hunting mechanism;
t7. if the current iteration number t is smaller than the maximum iteration number t max Then the DE-WOA algorithm is entered again to carry out optimal parameter selection; otherwise output whenThe number of neurons, learning rate, weight and bias of the previous best; and retraining the BP neural network by using the parameter configurations to obtain a DE-WOA-BP neural network model.
Further, the differential evolution algorithm in step T2 randomly selects three different parent individuals A, B, C to form a difference vector, and combines each element in the difference vector with the corresponding element of the parent individual to generate a child individual, and the generation formula is as follows:
X new ＝A+F·(B-C)
wherein X is new Is the offspring individual generated, F is the differential scaling factor constant;
and calculating the fitness of the individuals by using an individual fitness function F (x), comparing the fitness of the offspring individuals and the parent individuals, and selecting the individuals with higher fitness as the individuals in the next generation population.
Further, in step T3, the coefficient a=2a·r-a, r is a random number in the range of [0,1], a decreasing linearly from 2 to 0 in each iteration.
3. Advantageous effects
Compared with the prior art, the technical scheme provided by the invention has the following remarkable effects:
the differential evolution algorithm applied in the invention takes global search as a main strategy, and searches a parameter space by generating and evolving candidate solutions. This global search capability helps to avoid the BP neural network from falling into a locally optimal solution, as it can explore different regions of the parameter space, finding better combinations of weights and biases. Meanwhile, the whale algorithm is inspired by whale social behaviors, and can help the whale algorithm to maintain diversity in a search space by combining diversity of a differential evolution algorithm, so that premature convergence is prevented. This helps to search the solution space more comprehensively, finding the potentially best solution. The two algorithms can be combined to adjust the super parameters of the BP neural network, such as the number of neurons, the learning rate, the weight, the bias and the like. The DE-WOA algorithm can be used for initializing weights better to reduce the gradient disappearance problem in the training process, can also accelerate the network convergence speed and the fault recognition precision, overcomes the problems of low artificial estimation accuracy and poor economic benefit of an alloy batching scheme, improves the component hit rate and the steel product stability in the converter steelmaking process, effectively reduces the alloying cost, and has good application scenes.
Drawings
FIG. 1 is a training flow chart of a DE-WOA-BP neural network model in the invention;
FIG. 2 is a graph of training loss and test loss of the DE-WOA-BP neural network model of the present invention.
Detailed Description
For a further understanding of the present invention, the present invention will be described in detail with reference to the drawings and examples.
Example 1
And (3) smelting AISI1006B series steel grades in a certain steel plant, wherein ferrosilicon and low-carbon ferromanganese are required to be added in the tapping process so as to meet the requirement of the element content of the finished steel. In the actual production process, the alloy addition is estimated by operators through experience, the alloy types are selected to be not fixed, the alloy addition tends to be conservative and inaccurate, the later alloy addition is often excessive, the fluctuation of molten steel components of converter tapping is large, and the production rhythm and the economic benefit are further affected. To solve this series of problems, the present invention uses a DE-WOA-BP neural network model to determine the alloy addition.
The method comprises the steps of collecting 526 groups of production data of a converter of the factory 120t, and respectively processing missing data, repeated data and abnormal data of the production data, wherein a judgment formula of the abnormal data is as follows:
wherein x is i Ith data representing a specified field in the converter production environment dataset;a data average representing specified fields in the dataset; n represents the number of data in a specified field in the dataset.
After data processing, 479 sets of effective production data are obtained. To normalize the partial field data in the 479 set of data to interval 0,1 for different variables to have the same metric, the specific data reduction formula is:
wherein x is i ' represents normalized ith data; x is x i The ith original data representing a designated field in the converter production environment data set; min (x) and max (x) respectively represent the minimum value and the maximum value of data of a designated field in the converter production environment data set.
According to the result of the converter steelmaking reaction mechanism and Pearson correlation analysis, determining the following technological parameters with obvious influence on the addition amount of the alloy as the input amount of a model:
the molten iron carbon content, the molten iron silicon content, the molten iron manganese content, the molten iron phosphorus content, the molten iron sulfur content, the molten iron temperature, the molten iron loading amount, the scrap steel loading amount, the pig iron loading amount, the terminal temperature, the terminal carbon content and the terminal phosphorus content.
Newly building a BP neural network model of a network layer (input layer, hidden layer, output layer), wherein the input layer uses a fully connected layer with 12 neurons and uses a ReLU as an activation function; the hidden layer uses a fully connected layer with 6 neurons and uses ReLU as an activation function; the output layer uses a fully connected layer with 1 neuron and uses a linear activation function. The obtained preprocessed converter production environment data set is divided into two groups, wherein one group is used for training the DE-WOA-BP neural network model, and the other group is used for checking the prediction effect of the trained DE-WOA-BP neural network model.
And (3) iteratively establishing the neuron number, the learning rate, the weight and the bias parameters of the BP neural network model by using a DE-WOA algorithm and an fitness function, evaluating the performance of the parameter combination according to the fitness function return value, and applying the parameter combination with the minimum return value as the optimal parameter combination to the BP neural network to obtain the DE-WOA-BP neural network model. Training and testing a DE-WOA-BP neural network model, collecting real-time process variables of a converter smelting process required by the DE-WOA-BP neural network model, predicting the process variables by using the DE-WOA-BP neural network model, and calculating the type and weight of alloy to be added in the converter smelting process; alloy addition is carried out according to the predicted result of the DE-WOA-BP neural network model, tapping is finished, and data are stored in a database of the DE-WOA-BP neural network model.
For the problems of easy local optimum sinking, exploration and development incompatibility existing in the whale algorithm, after the differential evolution algorithm is introduced, individual differences among populations can be reduced, individuals are forced to leave the current search area, and the sinking of local optimum is avoided. After the differential evolution algorithm and the whale algorithm are fused, an optimized DE-WOA algorithm can be obtained, and the specific steps are as follows:
t1, setting the size of the population to be 30, and simultaneously carrying out random initialization on the population, and calculating to obtain the individual fitness f and the average fitness aver of the population. The optimization objective is to minimize the error, and the individual fitness function is set to be the inverse of the root mean square error, i.e
And T2, if f is less than or equal to aver, introducing a differential evolution algorithm to optimize a whale algorithm, sequentially performing mutation, crossover and selection operations, and applying a global search strategy to avoid trapping in local optimization. The differential evolution algorithm will randomly select three different individuals (parent individuals, assumed to be A, B, C) to construct a difference vector, combine each element in the difference vector with the corresponding element of the parent individual to produce a child individual, and generate the following formula:
X new ＝A+F·(B-C)
wherein X is new Is the offspring individual generated and F is the differential scaling factor constant.
Next, the fitness of the individuals is calculated by using the fitness function F (x), fitness of the offspring individuals and the parent individuals is compared, and the individuals with higher fitness are selected as the individuals in the next generation population.
And T3, if f > aver, replacing the current individual as the optimal individual in the population, and calculating a coefficient A and a random number p. Where the coefficient a=2a·r-a, r is a random number in the range of [0,1], a decreasing linearly from 2 to 0 in each iteration.
T4 if p is greater than or equal to 0.5, a spiral update mechanism is applied.
T5. if p <0.5 and |a| <1, then a shrink wrap mechanism is applied.
T6. if p <0.5 and |a| is not less than 1, a random hunting mechanism is applied.
T7. if the current iteration number t is smaller than the maximum iteration number t max Then the DE-WOA algorithm is entered again to carry out optimal parameter selection; otherwise, outputting the current optimal result, namely the optimal neuron number, the learning rate, the weight and the bias; and retraining the BP neural network by using the parameter configurations to obtain a DE-WOA-BP neural network model.
After optimizing parameters by a DE-WOA algorithm, applying the optimized parameters to a BP neural network, training to obtain a DE-WOA-BP neural network model, wherein the change of loss values in the training process is shown in figure 2. To evaluate the accuracy of the DE-WOA-BP neural network model, a root mean square error RMSE and a decision coefficient R are selected 2 As a performance evaluation index. RMSE is a measure of the accuracy of a model by calculating the difference between the predicted value and the actual observed value. The smaller the value of RMSE, the better the predictive performance of the model, as it measures the average degree of difference between the predicted value and the actual observed value. One advantage of RMSE over other error indicators is that it is insensitive to outliers because the difference is squared and then summed. R is R 2 One of the indicators used to measure the performance of the regression model is a statistic used to measure the degree of fit of the model to the observed data. After the DE-WOA-BP neural network model in the invention predicts and evaluates the data to be predicted, the obtained RMSE value is 2.5126 and R 2 Is 0.8942. From the evaluation result, the DE-WOA-BP neural network has good prediction performance and data interpretation capability, and can provide reference for on-site production.
The differential evolution algorithm applied in the invention takes global search as a main strategy, and searches a parameter space by generating and evolving candidate solutions. This global search capability helps to avoid the BP neural network from falling into a locally optimal solution, as it can explore different regions of the parameter space, finding better combinations of weights and biases. Meanwhile, the whale algorithm is inspired by whale social behaviors, and can help the whale algorithm to maintain diversity in a search space by combining diversity of a differential evolution algorithm, so that premature convergence is prevented. This helps to search the solution space more comprehensively, finding the potentially best solution. The two algorithms can be combined to adjust the super parameters of the BP neural network, such as the number of neurons, the learning rate, the weight, the bias and the like. The DE-WOA algorithm can be used for initializing weights better to reduce the gradient disappearance problem in the training process, can also accelerate the network convergence speed and the fault recognition precision, overcomes the problems of low artificial estimation accuracy and poor economic benefit of an alloy batching scheme, improves the component hit rate and the steel product stability in the converter steelmaking process, effectively reduces the alloying cost, and has good application scenes.
The invention and its embodiments have been described above by way of illustration and not limitation, and the invention is illustrated in the accompanying drawings and described in the drawings in which the actual structure is not limited thereto. Therefore, if one of ordinary skill in the art is informed by this disclosure, the structural mode and the embodiments similar to the technical scheme are not creatively designed without departing from the gist of the present invention.
Claims (7)
1. A method for determining the addition of a converter alloy by combining differential evolution and whale algorithm, which is characterized by comprising the following steps:
s1, collecting converter production environment data, and constructing a database for training a neural network model and storing model prediction results after manual correction;
s2, according to a metallurgical principle, establishing a process variable which influences the alloy addition amount in the tapping process of the converter in a database and taking the process variable as the input amount of a neural network model;
s3, preprocessing the data, creating a BP neural network model of a three-layer network layer, and using the obtained preprocessed converter production environment data set for training the neural network model;
s4, iterating the neuron number, the learning rate, the weight and the bias parameters of the BP neural network model established in the step S3 by using a DE-WOA algorithm and an fitness function, evaluating the performance of the parameter combination according to the fitness function return value, and applying the parameter combination with the minimum return value as the optimal parameter combination to the BP neural network to obtain the DE-WOA-BP neural network model;
s5, training and testing the DE-WOA-BP neural network model;
s6, collecting real-time process variables of the required converter smelting process, predicting the process variables by using a DE-WOA-BP neural network model, and calculating the type and weight of the alloy to be added in the converter smelting process; and (3) carrying out alloy addition according to the prediction result, and after tapping, storing the data into a database of the DE-WOA-BP neural network model.
2. The method for determining the addition amount of the converter alloy by combining differential evolution and whale algorithm according to claim 1, wherein the method comprises the following steps of: in the step S2, the input amount of the neural network model after analysis according to the correlation between the converter steelmaking reaction mechanism and Pearson specifically includes:
molten iron carbon content, molten iron silicon content, molten iron manganese content, molten iron phosphorus content, molten iron sulfur content, molten iron temperature, molten iron loading, scrap steel loading, pig iron loading, terminal temperature, terminal carbon content, terminal phosphorus content and alloy weight.
3. The method for determining the addition amount of the converter alloy by combining differential evolution and whale algorithm according to claim 2, wherein the method comprises the following steps of: the data preprocessing operation in step S3 includes data cleansing and data reduction.
4. A method for determining converter alloy additions according to any of claims 1-3, which is characterized by the following steps: the BP neural network model newly built in the step S3 comprises an input layer, an implicit layer and an output layer, wherein the input layer uses a full-connection layer with 12 neurons and uses a ReLU as an activation function; the hidden layer uses a fully connected layer with 6 neurons and uses ReLU as an activation function; the output layer uses a fully connected layer with 1 neuron and uses a linear activation function.
5. The method for determining the addition amount of the converter alloy by combining differential evolution and whale algorithm according to claim 4, wherein the method comprises the following steps of: the specific steps for obtaining the DE-WOA-BP neural network model in the step S4 are as follows:
t1, setting the size of a whale population, randomly initializing the population, calculating individual fitness f and average fitness aver of the population, optimizing the objective to minimize error, and setting an individual fitness function as the inverse of root mean square error;
t2, if f is less than or equal to aver, introducing a differential evolution algorithm to optimize a whale algorithm, sequentially performing mutation, crossover and selection operations, and applying a global search strategy to avoid trapping in local optimization;
if f > aver, replacing the current individual as the optimal individual in the population, and calculating a coefficient A and a random number p;
t4 if p is more than or equal to 0.5, applying a spiral update mechanism;
t5. if p <0.5 and |a| <1, applying a shrink wrap mechanism;
t6. if p <0.5 and |a| is not less than 1, applying a random hunting mechanism;
t7. if the current iteration number t is smaller than the maximum iteration number t max Then the DE-WOA algorithm is entered again to carry out optimal parameter selection; otherwise, outputting the current optimal neuron number, learning rate, weight and bias; and retraining the BP neural network by using the parameter configurations to obtain a DE-WOA-BP neural network model.
6. The method for determining the addition amount of the converter alloy by combining differential evolution and whale algorithm according to claim 5, wherein the method comprises the following steps of: in step T2, the differential evolution algorithm randomly selects three different parent individuals A, B, C to form a differential vector, and combines each element in the differential vector with the corresponding element of the parent individual to generate a child individual, where the generation formula is as follows:
X new ＝A+F·(B-C)
wherein X is new Is the offspring individual generated, F is the differential scaling factor constant;
and calculating the fitness of the individuals by using an individual fitness function F (x), comparing the fitness of the offspring individuals and the parent individuals, and selecting the individuals with higher fitness as the individuals in the next generation population.
7. The method for determining the addition amount of the converter alloy by combining differential evolution and whale algorithm according to claim 6, wherein the method comprises the following steps of: in step T3, the coefficient a=2a·r-a, r is a random number in the range of [0,1], a decreasing linearly from 2 to 0 in each iteration.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410050745.2A CN117877617A (en) | 2024-01-12 | 2024-01-12 | Method for determining addition amount of converter alloy by combining differential evolution and whale algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410050745.2A CN117877617A (en) | 2024-01-12 | 2024-01-12 | Method for determining addition amount of converter alloy by combining differential evolution and whale algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117877617A true CN117877617A (en) | 2024-04-12 |
Family
ID=90592936
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410050745.2A Pending CN117877617A (en) | 2024-01-12 | 2024-01-12 | Method for determining addition amount of converter alloy by combining differential evolution and whale algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117877617A (en) |
-
2024
- 2024-01-12 CN CN202410050745.2A patent/CN117877617A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106096788B (en) | Converter steelmaking process cost control method and system based on PSO _ ELM neural network | |
| CN114611844B (en) | Method and system for determining alloy addition amount in converter tapping process | |
| CN113128124B (en) | Multi-grade C-Mn steel mechanical property prediction method based on improved neural network | |
| Wang et al. | Strip hardness prediction in continuous annealing using multiobjective sparse nonlinear ensemble learning with evolutionary feature selection | |
| CN106650944A (en) | Metallurgical enterprise converter gas scheduling method based on knowledge | |
| CN113642220A (en) | Ship welding process optimization method based on RBF and MOPSO | |
| CN113869795B (en) | Long-term scheduling method for industrial byproduct gas system | |
| Ordieres-Meré et al. | Comparison of models created for the prediction of the mechanical properties of galvanized steel coils | |
| CN117877617A (en) | Method for determining addition amount of converter alloy by combining differential evolution and whale algorithm | |
| CN111126550B (en) | Neural network molten steel temperature forecasting method based on Monte Carlo method | |
| CN108171381A (en) | A kind of blast furnace CO utilization rates chaos weighing first order local prediction method and system | |
| CN117093868A (en) | Converter endpoint prediction method and device based on multitask learning | |
| CN116484745A (en) | Design method of hot-rolled low-alloy steel for physical metallurgy guidance industry big data mining | |
| CN115456264A (en) | Method for predicting end point carbon content and end point temperature of small and medium-sized converters | |
| Yang et al. | Genetic algorithms and hybrid neural network modelling for aluminium stress—strain prediction | |
| Pengtao | Based on adam optimization algorithm: Neural network model for auto steel performance prediction | |
| Bahadır et al. | Robot selection for warehouses | |
| CN111861041A (en) | Method for predicting dynamic recrystallization type flowing stress of Nb microalloyed steel | |
| Gaffour et al. | ‘Symbiotic’data-driven modelling for the accurate prediction of mechanical properties of alloy steels | |
| Liu et al. | Establishment and Application of Steel Composition Prediction Model Based on t-Distributed Stochastic Neighbor Embedding (t-SNE) Dimensionality Reduction Algorithm | |
| Li et al. | Application of rough set theory and artificial neural network for load forecasting | |
| CN114528770A (en) | Multi-objective optimization method for copper flash smelting process | |
| Huang et al. | Prediction of alloy yield based on optimized BP neural network | |
| CN103135444A (en) | Steel production energy consumption immune prediction control model | |
| Han et al. | A multi-objective optimization model for alloy addition in BOS process based on ESN and modified MOPSO |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination |