Steel Temperature Prediction Model for the Tempering Process of the Whole-Element Driving Fusion Mechanism, Its Construction and Application

Abstract

The tempering process steel temperature prediction model for the all-factor driven fusion mechanism, its construction and application belong to the field of optimization control of tempering furnaces in iron and steel metallurgy, and are used to make the steel temperature prediction model interpretable, physically meaningful and extrapolable, and have higher accuracy. The key point is to obtain a training data set; among them, the input data of the tempering process steel temperature prediction model in the training data set is the operation data of the tempering furnace, and the output data is the overall heat absorption rate; the tempering process steel temperature prediction model is trained by the training set. Among them, the overall heat absorption rate is obtained through steps such as establishing an identification objective function with the overall heat absorption rate as the core and using an improved GSK algorithm to identify the objective function. The present invention enhances the accuracy and robustness when facing new data, and effectively solves the ill-posedness challenge in the solution of inverse problems in the field of steel temperature prediction.

Description

Technical Field

[0001] The present invention belongs to the field of optimized control of tempering furnaces in iron and steel metallurgy, and particularly relates to a steel temperature prediction model for the tempering process based on the integration mechanism of all-element drive, and its construction and application. Background Art

[0002] As a solid cornerstone of the national economy and industrial development, the importance of the iron and steel industry is self-evident. Facing multiple severe challenges such as high energy consumption, high emissions, and high pollution, iron and steel enterprises have always adhered to the innovation concept and are committed to making breakthroughs in the efficient utilization of raw materials, the innovation of production processes, and the adoption of environmental protection technologies. From ore mining to ironmaking, steelmaking, and rolling, the iron and steel industrial chain covers multiple key links, and each link highly depends on the stable operation of thermal equipment. Therefore, the refined optimization and adjustment of the operating parameters of thermal equipment and the further improvement of process models are of great significance for the iron and steel industry to promote energy conservation and emission reduction and improve comprehensive benefits.

[0003] The tempering furnace is an important heat treatment equipment after quenching, and its main function is to heat the steel plate to a specific preset temperature. However, this heating process is accompanied by significant energy consumption problems. Therefore, improving the automation level of the tempering furnace is of extremely important significance for achieving the goal of energy conservation and emission reduction. The temperature distribution state of the steel plate inside the tempering furnace is not only a key index for evaluating the quality of the heating process but also an important basis for optimizing the temperature control strategy of the tempering furnace. The heating process of the steel plate is extremely complex, showing characteristics of non-linearity, large time delay, strong coupling, and involving multiple heat transfer mechanisms. In addition, the internal environment of the tempering furnace is harsh, with problems of high temperature and dense dust, and there are many challenges in directly and accurately measuring the temperature of the steel plate. In view of this, establishing an accurate steel plate temperature prediction model to estimate the temperature of the steel plate in the furnace has become the focus of current research.

[0004] The mathematical models for steel plate temperature prediction mainly fall into two categories: data-driven models and mechanism models. Data-driven models focus on widely collecting and deeply analyzing industrial data, aiming to discover the implicit relationships between data and build an accurate temperature prediction model based on this. In contrast, mechanism models use the finite element method or the finite difference method to discretize the steel plate and strictly define the heat transfer equation and boundary conditions according to the basic principles of heat transfer, so as to build a theoretically supported temperature prediction model. In addition, there is also a hybrid model that effectively combines the advantages of mechanism models and data-driven models.

[0005] The Chinese patent "CN114015863B An Algorithm for Self-Correcting the Steel Billet Heating Model" provides an algorithm for automatically correcting the steel billet heating model. This patent uses the overall absorptivity method to establish a two-dimensional steel billet heating model, determines the preliminary overall absorptivity through coupling experiments, and calculates the non-uniformity coefficient of the temperature distribution of the steel billet along the furnace length direction; uses the heat balance method to calculate the heat absorbed by the steel billet during this period, and uses the steel billet heating model to calculate the total heat absorbed by all the steel billets in the furnace during this period. This method predicts the change of the internal temperature distribution of the steel billet during the heating process through the mathematical model of the heating furnace, controls the combustion process of the heating furnace using the optimized furnace temperature setting value, and continuously corrects it in real time according to the on-site operation conditions.

[0006] The Chinese patent "CN113849020B A Method and Device for Designing the Heating Curve of Steel Billets Based on Artificial Intelligence Algorithm", this method divides each control section of the heating furnace into multiple subsections; obtains the historical heating data corresponding to each subsection; uses the historical heating data corresponding to each subsection to train a preset neural network model respectively to obtain the steel billet temperature prediction model corresponding to each subsection; according to the current values of the temperature influence factors corresponding to each subsection, for each subsection, uses the steel billet temperature prediction model corresponding to the current subsection to predict the predicted value of the outlet temperature at the end of the subsection corresponding to this subsection; fits the predicted values of the outlet temperatures at the ends of the subsections corresponding to each other to obtain the heating curve of the steel billet.

[0007] The Chinese patent "CN113821984B A Method for Calculating the Temperature of Steel Billets in a Heating Furnace Based on a Time Domain Convolution Model", this patent first establishes a mechanism model for two-dimensional heat transfer of steel billets based on the basic principles of heat transfer, and calculates the temperature distribution of the steel billets in real time; then, collects the steel grade, size, time in the furnace, furnace temperature of the steel billets, and the temperature of the steel billets calculated by the mechanism model, forms a characteristic data set with the temperature of the steel billets measured by the infrared thermometer and performs data preprocessing. Finally, based on the time domain convolution model, a correction model for the temperature of the steel billets is established, the model is trained using the data set, and the trained model is used to correct the temperature of the steel billets in the furnace.

[0008] However, the technical solution described in the Chinese patent "CN113849020B A Method and Device for Designing the Heating Curve of Steel Billets Based on Artificial Intelligence Algorithm" completely abandons the mechanism knowledge of the heat transfer process. Therefore, the strong nonlinearity, time delay, and high dimensionality of the heating process in the furnace all have certain adverse effects on the modeling process and model performance, and the data-driven model has low interpretability, physical meaning, and extrapolability.

[0009] The technical solution described in the Chinese patent "CN114015863B A Self-Correcting Algorithm for Steel Billet Heating Model" requires the prior determination of unknown heat transfer parameters in the equation. In the mechanism model of steel plate heating, the heat transfer parameters are determined based on on-site thermocouple embedding experiments. By means of thermocouple embedding experiments, the workload of parameter acquisition is large, the debugging time is long, and it is easy to affect steel plate production.

[0010] For the technical solution described in the Chinese patent "CN113821984B A Method for Calculating the Temperature of Steel Billet in a Heating Furnace Based on a Time Domain Convolution Model", the input data volume is insufficient, only including measured furnace temperature, measured steel temperature, and data related to the steel billet, while ignoring important influencing factors such as the operating state of the burner, fault state, and the flow rates of fuel and combustion-supporting air. Therefore, it cannot meet the requirements of multiple steel types and multiple working conditions. And it only corrects the temperature of the steel billet at the location of the infrared thermometer, while ignoring the heating-up process of the steel billet. Summary of the Invention

[0011] To solve the above problems, the present invention effectively reflects the process characteristics and laws through a mechanism model, and has good interpretability and extrapolability. The overall heat absorption rate output by the all-factor-driven recurrent neural network does not require thermocouple embedding experiments, and at the same time overcomes the derivation of empirical formulas, and the obtained overall heat absorption rate is more accurate.

[0012] In a first aspect, a method for constructing a steel temperature prediction model for the tempering process based on all-factor-driven fusion mechanism according to some embodiments of the present application includes

[0013] Obtaining a training data set; wherein, the input data of the steel temperature prediction model for the tempering process in the training data set is the operation data of the tempering furnace, and the output data is the overall heat absorption rate;

[0014] Training the steel temperature prediction model for the tempering process with the training set.

[0015] According to the method for constructing a steel temperature prediction model for the tempering process based on all-factor-driven fusion mechanism according to some embodiments of the present application, the training data set is obtained based on the following method:

[0016] S3.1. Initialize a population of size NP, and the first individual x1 is randomly initialized to generate variables within its specified range;

[0017] S3.2. Obtain the x sequence through the Chebyshev mapping as the initialized population;

[0018] S3.3. Sort all individuals in the current population in ascending order according to the objective function value: x1, x2,..., x NP , update the evaluation times F E =NP, where x1 is the individual with the optimal solution of the objective function value, and x NP is the individual with the worst solution of the objective function value;

[0019] S3.4. (1) When the random number rand is greater than the knowledge ratio K r do not update the individual, calculate the fitness value of the objective function, and update the evaluation times F E = F E+ NP, and update the population; (2) When the random number rand is not greater than the knowledge ratio K r then judge:

[0020] (2-1) When the random number rand is not less than the number of dimensions N updated by the primary profit sharing knowledge scheme junior update the individual using the formula in the primary knowledge sharing stage calculate the fitness value of the objective function, and update the evaluation times F E = F E+ NP, and update the population;

[0021] (2-2) When the random number rand is less than the number of dimensions N updated by the primary profit sharing knowledge scheme junior update the individual using the formula in the advanced knowledge sharing stage calculate the fitness value of the objective function, and update the evaluation times F E = F E+ NP, and update the population;

[0022] S3.5. Use the greedy selection method to decide whether to use the old individual or the new individual in the next iteration;

[0023] S3.6. Substitute the overall heat absorption rate of each furnace zone corresponding to each dimension of the selected individual into the two-dimensional unsteady heat transfer model of the tempering furnace to calculate the outgoing temperature of the j-th steel plate measured when leaving the furnace and the upper surface temperature of the j-th steel plate measured by the in-furnace pyrometer

[0024] S3.7. Substitute the outgoing temperature of the j-th steel plate measured when leaving the furnace and the upper surface temperature of the j-th steel plate measured by the in-furnace pyrometer and into the identification objective function to solve the optimal overall heat absorption rate of the furnace zone;

[0025] S3.8. After updating the global optimum using the greedy selection method, judge whether the maximum number of iterations is reached. If the maximum number of iterations is reached, output the optimal overall heat absorption rate of each furnace zone and the operation data of the tempering furnace for constructing the training dataset; if the maximum number of iterations is not reached, return to step S3.3.

[0026] According to the method for constructing a steel temperature prediction model during the tempering process based on the full-factor drive fusion mechanism of some embodiments of the present application, wherein:

[0027] In step S3.1, initialize a population of size NP. The first individual x1 is randomly initialized with variables within its specified range as shown in the following equation:

[0028]

[0029] where x i is the i-th individual, i = 1, 2, …, NP, and x i,j represents the j-th dimension of the individual x i ; and are the lower and upper bounds of the j-th decision variable respectively, and is a random number within the range [0, 1];

[0030] In step S3.2, obtain the x sequence through the Chebyshev mapping as the initialized population, as shown in the following equation:

[0031] x i+1 = cos(a * cos -1 (x i )), a = 4 (11)

[0032] where x i+1 is the (i + 1)-th individual, and a represents the order;

[0033] In step S3.4, update the formula in the primary knowledge sharing stage:

[0034]

[0035] where represents the j-th dimension of the individual ; represents the j-th dimension of the individual ; K f represents the knowledge factor; represents the j-th dimension of the individual , Guass(0, 1) represents a Gaussian distribution with a mean of 0 and a variance of 1, represents the j-th dimension of the random individual;

[0036] In step S3.4, update the formula in the advanced knowledge sharing stage:

[0037]

[0038] where represents the j-th dimension of the best individual , represents the j-th dimension of the worst individual , represents the j-th dimension of the random individual;

[0039] In step S3.5, a greedy selection method is adopted to determine whether to use the old individual or the new individual in the next iteration, as shown in the following formula:

[0040]

[0041] In the formula, represents the selected individual, is the solution generated in the primary and advanced shared knowledge acquisition stages, is the i-th solution in the current population.

[0042] According to the method for constructing a steel temperature prediction model during the tempering process of the all-factor-driven fusion mechanism according to some embodiments of the present application, the tempering furnace operation data includes any one or combination of the burner duty ratio, burner fault status, fuel supply and air-fuel ratio, gas preheating temperature, steel plate speed, steel plate position, furnace zone temperature, measured length of the steel plate, measured width, measured thickness, steel plate number, steel plate type, steel plate inlet temperature, surface temperature of the steel plate measured by the in-furnace pyrometer, steel plate outlet temperature, steel plate running speed, furnace length, heating time, holding time, and temperature measured by the thermocouple in each furnace zone.

[0043] According to the method for constructing a steel temperature prediction model during the tempering process of the all-factor-driven fusion mechanism according to some embodiments of the present application, the tempering furnace operation data includes the burner duty ratio, burner fault status, fuel supply and air-fuel ratio, gas preheating temperature, steel plate speed, steel plate position, and furnace zone temperature.

[0044] According to the method for constructing a steel temperature prediction model during the tempering process of the all-factor-driven fusion mechanism according to some embodiments of the present application, the mathematical expression of the identification objective function is as follows:

[0045]

[0046] In the formula, J represents the objective function, N represents the number of steel plates, represents the outlet temperature of the j-th steel plate measured at the outlet; represents the upper surface temperature of the j-th steel plate measured by the in-furnace pyrometer; represents the outlet temperature of the j-th steel plate calculated by the heat transfer model of the tempering furnace; represents the upper surface temperature of the j-th steel plate when it is transported to the position where the in-furnace pyrometer is located, calculated by the heat transfer model of the tempering furnace; W represents the combined vector of the overall heat absorption rates at different positions along the furnace length, represents the m-th overall heat absorption rate along the furnace length, and λ1 and λ2 represent the regularization coefficients.

[0047] Method for constructing a steel temperature prediction model for the tempering process of the full-factor drive fusion mechanism according to some embodiments of the present application. The two-dimensional unsteady heat transfer model of the tempering furnace is established based on the following method:

[0048] S1.2. Establish a two-dimensional unsteady heat conduction equation, and the specific mathematical description is as follows:

[0049]

[0050] In the formula, ρ represents the density of the steel plate; C p represents the specific heat of the steel plate; T represents the temperature of the steel plate; τ represents time; λ represents the thermal conductivity of the steel plate; represents the differential along the length direction of the steel plate, represents the differential in the thickness direction of the steel plate;

[0051] S1.3. Establish boundary conditions. The mathematical description of the heat flux density on the steel plate surface is:

[0052]

[0053] In the formula, q represents the heat flux density on the steel plate surface; σ represents the Boltzmann constant; represents the total heat absorption rate; T f represents the furnace temperature, T s represents the temperature of the steel plate surface;

[0054] S1.4. According to the position of any steel plate in the furnace changing with time, transform the boundary conditions where the steel plate is located into a time-varying temperature field problem. Among them, for any steel plate i, the position x of the tail of the steel plate in the furnace i is determined by the moving speed of the steel plate and is expressed as follows:

[0055]

[0056] In the formula, x i (v(t),t) represents the position of the tail of the steel plate in the furnace, t represents the time variable, t f is the time of the steel plate in the furnace, dτ represents the time differential, and ν(t) is the moving speed of the steel plate at time t;

[0057] S1.5. Calculate the boundary conditions, which are expressed as follows:

[0058]

[0059] In the formula, T(τ,x i ,y) represents the temperature of the steel plate, x i represents the position of the tail of the steel plate, x represents the coordinate axis along the length direction of the steel plate, y represents the coordinate axis in the thickness direction of the steel plate, b represents the thickness of the steel plate, a represents the length of the steel plate, q h and q trespectively represent the heat flux density on the front side and the heat flux density on the rear side of the steel plate, q u and q l respectively represent the heat flux density on the upper surface and the heat flux density on the lower surface of the steel plate;

[0060] S1.6. Discretize the two-dimensional unsteady heat conduction equation and boundary conditions by using the finite difference method, so as to iterate the heat transfer model of the two-dimensional unsteady tempering furnace and output the relevant temperatures at each moment. Among them, the relevant temperatures include the outlet temperature of the j-th steel plate measured at the outlet and the upper surface temperature of the j-th steel plate measured by the high-temperature thermometer in the furnace as well as the internal temperature of the steel plate.

[0061] According to the method for constructing a steel temperature prediction model of the full-factor drive fusion mechanism according to some embodiments of the present application, the model is a recurrent neural network, and the input vector x t =[x 1,t , x 2,t ,..., x M,t represents the state of the tempering furnace operation data at time t, M represents the number of types of tempering furnace operation data, and the output vector y t represents the estimated value of the overall heat absorption rate along the furnace length direction at time t;

[0062] For any given time step t, the RNN recursively calculates the new hidden state h t and the predicted value y t :

[0063] h t =tanh(W hx x t +W hh h t-1 +b h ) (17)

[0064] y t =W hy h t +b y (18)

[0065] In the formula, W hx ∈R m×n , W hh ∈R m×m represent weight matrices, W hy represents the weight matrix of the output layer, x t represents the input layer, h t-1 represents the hidden layer, n represents the number of input features, b h ∈R m and b y∈R represents the bias terms for the hidden layer and the output layer respectively. The tanh function is used to introduce non-linearity, and m represents the number of neurons in the hidden layer;

[0066] The model training includes

[0067] S4.1. Initialize the parameters, randomly initialize all weights W hx , W hh , W hy and the bias terms b h , b y ;

[0068] S4.2. Forward propagation, given a batch of training samples where is a series of consecutive observations, is the corresponding total heat absorption rate, and perform forward propagation to obtain the predicted value In the formula, N represents the number of groups of input data, represents the output value of the T-th dimension in the training sample, represents the predicted value of the T-th dimension during the training process, and T represents the dimension of the output layer;

[0069] S4.3. Calculate the loss, which is expressed as follows:

[0070]

[0071] In the formula, L represents the loss function;

[0072] S4.4. Use the gradient descent method to adjust the parameters to minimize the loss function for backpropagation, including: calculate the gradient ▽L with respect to W hx , W hh , W hy , b h , b y , and update the parameters according to the learning rate η, which is expressed as follows:

[0073]

[0074]

[0075]

[0076] In the formula, represents the partial derivative of the weight, and L represents the loss function;

[0077] S4.5. Loop the training until the stop condition is met.

[0078] In a second aspect, the tempering process steel temperature prediction model of the all-factor drive fusion mechanism according to some embodiments of the present application is constructed by any one of the methods in the first aspect.

[0079] In a third aspect, a method for predicting the steel temperature during the tempering process of the full-factor drive fusion mechanism according to some embodiments of the present application is characterized in that it is implemented by the steel temperature prediction model for the tempering process constructed by the method described in the second aspect;

[0080] The method includes:

[0081] Input the operation data of the tempering furnace into the steel temperature prediction model for the tempering process;

[0082] The steel temperature prediction model for the tempering process outputs the overall heat absorption rate of each furnace zone;

[0083] According to the overall heat absorption rate of each furnace zone, calculate the steel temperature during the tempering process through a two-dimensional unsteady heat transfer model of the tempering furnace.

[0084] In a third aspect, an embodiment of the present application further provides an electronic device, which includes: one or more processors, a memory, and one or more programs; wherein, the one or more programs are stored in the memory, and the one or more programs include instructions, when the instructions are executed by the electronic device, the electronic device is enabled to execute the first, third aspects and any possible technical solutions thereof.

[0085] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, which includes a computer program, when the computer program runs on an electronic device, the electronic device is enabled to execute the first aspect and the first, third aspects and any possible technical solutions thereof.

[0086] Advantages of the present invention: Compared with the prior art, the steel temperature prediction model constructed by the technical solution proposed by the present invention integrates heat transfer formulas, has better interpretability, physical meaning and extrapolability than data-driven models, and the fluctuation of the steel temperature output value is smaller. The overall heat absorption rate in the technical solution does not need to be determined in advance through coupling experiments, and the overall heat absorption rate is more in line with the actual situation than the value calculated by the empirical formula. Therefore, the model accuracy is higher than that of the mechanism model.

[0087] Specifically:

[0088] In the first aspect, the present invention lies in using the operation data of the full-factor tempering furnace as the input and the identified overall heat absorption rate as the output to train a recurrent neural network. The overall heat absorption rate of the tempering furnace is usually affected by a series of continuous operating parameters, the health status and operating status (duty cycle) of each burner, the furnace temperature of each temperature control zone, the fuel flow rate and pressure, the combustion air flow rate and pressure, etc., and these parameters change with time. The RNN can store historical information through the hidden layer and can effectively capture the time-dependent relationship in the input data, which is very important for understanding and predicting the heat absorption rate.

[0089] In the second aspect, a GSK algorithm combined with a Gaussian mutation strategy is proposed in the present invention to solve the problem of identifying the overall heat absorption rate of a tempering furnace. The initial population distribution is crucial for the convergence speed and accuracy. To maintain population diversity, the Chebyshev mapping is used to initialize the population. Compared with the random method, this makes the individual distribution more uniform, reduces overlap, and enhances the global search ability of the algorithm. In addition, the population diversity is further improved by introducing Gaussian mutation, which helps to explore a wider solution space, avoid falling into local optima, and thus increase the probability of finding superior solutions.

[0090] In the third aspect, an objective function with the overall heat absorption rate as the core is designed in the present invention. Based on a two-dimensional heat transfer model, the heating process of the steel plate is modeled, ensuring high precision and computational efficiency. A regularization strategy is applied in the objective function to optimize the structure, which not only prevents overfitting and improves the generalization ability of the model, but also enhances the accuracy and robustness when facing new data, effectively solving the ill-posedness challenge in inverse problem solving. Brief Description of the Drawings

[0091] Figure 1 Flowchart of the technical solution.

[0092] Figure 2 Schematic diagram of the heat absorption of a two-dimensional steel plate cross-section.

[0093] Figure 3 Flowchart of the improved GSK algorithm for identifying the overall heat absorption rate.

[0094] Figure 4 Schematic diagram of a recurrent neural network. Detailed Embodiments

[0095] The embodiments of the present application will be described in detail below with reference to the accompanying drawings. The examples of the embodiments are shown in the drawings. The present application provides a method and an electronic device. Among them, the method and the device are based on the same technical concept. Since the principles of the method and the device for solving problems are similar, the implementation of the device and the method can be referred to each other, and the repeated parts will not be described again.

[0096] As Figure 1 shown, the method for constructing a steel temperature prediction model for the tempering process based on the full-factor drive fusion mechanism proposed in the present invention includes the following steps:

[0097] S1: Construct a two-dimensional unsteady heat transfer model and boundary conditions of the tempering furnace.

[0098] S2: Establish an identification objective function with the overall heat absorption rate as the core.

[0099] S3: Use the improved GSK algorithm to identify the objective function.

[0100] S4: The operation data of the tempering furnace is used as the input, and the corresponding overall heat absorption rate is used as the output to train a recurrent neural network.

[0101] S5: The dynamically changing overall heat absorption rate output by the recurrent neural network is brought into the heat transfer model of the tempering furnace to obtain a steel plate temperature prediction model.

[0102] Figure 1 The main flow chart of the technical solution of the present invention is shown below. Taking the tempering heat treatment production line of a certain steel plant as an example:

[0103] The operation data of the tempering furnace includes: burner duty ratio, burner failure status, fuel supply and air-fuel ratio, gas preheating temperature, steel plate speed, steel plate position, furnace zone temperature, measured length of the steel plate, measured width, measured thickness, steel plate number, steel plate type, steel plate inlet temperature, surface temperature of the steel plate measured by the in-furnace pyrometer, steel plate outlet temperature, steel plate running speed, furnace length, heating time, holding time, temperature measured by thermocouples in each furnace zone

[0104] The steel plate specifications are 20mm×2000mm×12000mm, the roller table running speed is 0.2 - 20m / min, the steel plate inlet temperature is room temperature 20°C, and the process requires the surface temperature of the steel plate to reach 180°C when it leaves the furnace. The effective furnace length of the tempering furnace is 106m, there are 34 upper and lower furnace zones in total, 6 - 12 burners are installed in each furnace zone, the burner type is high-speed burner, and in-furnace pyrometers are installed in the middle and the tail of the tempering furnace for measuring the surface temperature of the steel plate. The algorithm termination condition is to reach the maximum number of iterations.

[0105] In step S1, the two-dimensional unsteady heat transfer model of the tempering furnace is shown in Figure 2 , and the construction steps are as follows:

[0106] S1.1: Construct a two-dimensional unsteady heat transfer model of the tempering furnace and boundary conditions. First, make the following assumptions to simplify the model.

[0107] a) Neglect the influence of the scale on the steel plate and the surface chemical change on the heat transfer condition.

[0108] b) Approximately consider that the ambient temperature where the steel plate is located is uniform within each temperature control zone;

[0109] c) Neglect the influence of heat conduction and radiation heat transfer between the steel plate and the furnace rolls;

[0110] d) The steel plate moves at a constant speed in the furnace, neglecting the heat conduction along the length direction of the steel plate;

[0111] e) For the convenience of calculation, express the convective heat transfer heat flux density in the form of the radiative heat transfer heat flux density, and convert it into a radiative treatment method, that is, use the overall heat absorption rate method to calculate the steel temperature.

[0112] f) It is considered that the medium temperature in the furnace is evenly distributed along the furnace width direction.

[0113] g) Considering the installation position and insertion depth of the thermocouple, it is approximately considered that the furnace gas temperature is equal to the furnace temperature.

[0114] Based on the above assumptions, a mathematical model for the steel plate heating process is established.

[0115] S1.2: Establish a two-dimensional unsteady heat conduction equation, and the specific mathematical description is as follows:

[0116]

[0117] In the formula, ρ represents the density of the steel plate, Kg / m 3 ; C p represents the specific heat of the steel plate, J / (kg●℃); T represents the temperature of the steel plate, ℃; τ represents time, s; λ represents the thermal conductivity of the steel plate, W / (m●℃); represents the differential along the length direction of the steel plate, represents the differential in the thickness direction of the steel plate.

[0118] S1.3: Establish boundary conditions, and the mathematical description of the heat flux density on the steel plate surface is:

[0119]

[0120] In the formula: q represents the heat flux density on the steel plate surface, W / m 2 ; σ represents the Boltzmann constant, 5.67●10 -8 W / (m 2 ●K 4 ); represents the total heat absorptivity; T f and T s represent the furnace temperature and the steel plate surface temperature respectively, K.

[0121] S1.4: Since the position of a specific steel plate in the furnace changes with time, a moving coordinate system is adopted, so that the boundary conditions where the steel plate is located are correspondingly transformed into a time-varying temperature field problem. Considering any steel plate i in the furnace, the position x of the tail of the steel plate in the furnace i is determined by the moving speed of the steel plate:

[0122]

[0123] In the formula: x i (v(t),t) represents the position of the tail of the steel plate in the furnace; t represents the time variable; t f is the time of the steel plate in the furnace; dτ represents the time differential; ν(t) is the moving speed of the steel plate at time t, m / min; t fThe time of the steel plate in the furnace, s.

[0124] S1.5: The boundary conditions corresponding to the steel plate tracking model are as follows:

[0125]

[0126] In the formula: T(τ, x i , y) represents, x i represents the position of the tail of the steel plate, x represents the coordinate axis along the length direction of the steel plate, y represents the coordinate axis in the thickness direction of the steel plate, b represents the thickness of the steel plate, a represents the length of the steel plate, q h and q t are respectively the heat flux densities on the front and rear sides of the steel plate, W / m 2 . q u and q l are respectively the heat flux densities on the upper and lower surfaces, W / m 2 .

[0127] S1.6: The finite difference method is used to discretize the heat conduction equation and boundary conditions of the steel plate. The calculation format of the numerical equation established by the finite difference method is an explicit difference format, and each node equation can be solved independently. Thus, the model is iterated to generate the relevant temperatures at each moment, including the tapping temperature of the jth steel plate measured at tapping and the upper surface temperature of the jth steel plate measured by the high-temperature thermometer in the furnace as well as the temperatures such as the internal temperature of the steel plate.

[0128] In step S2, an identification objective function with the overall heat absorption rate as the core is established, requiring the error between the model calculation value and the measured steel temperature value to be minimized. To solve the ill-posed problem, the regularization strategy is applied to the trimming of the objective function of this optimization problem. The purpose of this operation is to improve the generalization ability of the model while reducing the overfitting of the model to the data. The mathematical expression of the identification objective function is as follows:

[0129]

[0130] In the formula, J represents the objective function. In this embodiment, 20 steel plates are used. Among them, represents the upper surface temperature of the jth steel plate at tapping measured at tapping, °C; represents the upper surface temperature of the jth steel plate measured by the high-temperature thermometer in the furnace, °C; represents the tapping temperature of the jth steel plate calculated by the heat transfer model of the tempering furnace, °C; represents the upper surface temperature of the jth steel plate calculated by the heat transfer model of the tempering furnace when it is transported to the position of the high-temperature thermometer in the furnace, °C. W represents the combined vector of the overall heat absorption rates at different positions along the furnace length direction, It represents the total heat absorption rate of the m-th along the furnace length direction. Both λ1 and λ2 are regularization coefficients.

[0131] The identification and optimization process of the improved GSK algorithm mentioned in step S3 is as Figure 3 shown, and the running steps are as follows:

[0132] S3.1: This algorithm first initializes a population of size NP. The first individual x1 is a variable randomly initialized within its specified range, and the variable generation formula is as follows:

[0133]

[0134] where x i is the i-th individual, x i,j represents the j-th dimension of the individual x i , i = 1, 2, …, NP, and are the lower and upper bounds of the j-th decision variable respectively, and a random number is taken within the range of [0, 1]; rand() represents the random function

[0135] S3.2: The Chebyshev mapping has good ergodic uniformity and can improve the optimization speed of the algorithm. At the same time, it can generate relatively uniform initial values between [-1, 1]. Therefore, the Chebyshev mapping will be used to initialize the population, and the mapping expression is as follows:

[0136] x i+1 = cos(a * cos -1 (x i )), a = 4(11)

[0137] where x i+1 is the (i + 1)-th individual, and a represents the order.

[0138] The initial value of the individual x1 randomly generated in step S3.1 is iterated according to the above formula to generate an x sequence, i = i + 1; this x sequence serves as the initial population.

[0139] S3.3: Note that the number of dimensions updated or changed using the primary and advanced acquisition of shared knowledge schemes will depend on the knowledge rate K. This knowledge rate is used to control the amount of knowledge transferred between generations of other solutions using the primary and advanced acquisition of shared knowledge schemes. Therefore, it is necessary to calculate the number of dimensions to be changed using the primary stage and the number of dimensions to be changed using the advanced stage at the beginning of each generation.

[0140] Let N junior be the number of dimensions updated using the primary benefit sharing knowledge scheme, and the calculation formula is

[0141]

[0142] Where d is the problem scale; K is a positive integer greater than zero, representing the knowledge rate; g is the number of iterations; g max is the maximum number of iterations. In the advanced revenue sharing stage, the updated number of dimensions N senior is calculated by the following formula:

[0143] N senior = d - N junior (13)

[0144] Make the following judgments:

[0145] (1) When the random number rand is greater than the knowledge ratio K r , do not update the individual, calculate the fitness value of the objective function, and update the evaluation times F E = F E+ NP, and update the population; (2) When the random number rand is not greater than the knowledge ratio K r , make a judgment:

[0146] (2 - 1) When the random number rand is not less than the number of dimensions N updated by the primary revenue sharing knowledge scheme junior , update the individual using the formula in the primary knowledge sharing stage Calculate the fitness value of the objective function, and update the evaluation times F E = F E+ NP, and update the population;

[0147] (2 - 2) When the random number rand is less than the number of dimensions N updated by the primary revenue sharing knowledge scheme junior , update the individual using the formula in the advanced knowledge sharing stage Calculate the fitness value of the objective function, and update the evaluation times F E = F E+ NP, and update the population;

[0148] S3.4: The primary harvest sharing knowledge stage should follow the following steps:

[0149] Sort all individuals in the current population in ascending order according to the objective function value: x1, x2,..., x NP . Where x1 is the solution with the minimum objective function value (the best individual), and x NP is the solution with the worst objective function value (the worst individual). Then, for each solution x i , the following operations should be selected: (a) The better solution x i-1 nearest to the current solution and the worse solution x i+1 (the fitness of individual x i+1 is one rank lower than that of individual x i in the sorting) to form the knowledge gain source. (b) Randomly select another individual xr As a source of shared knowledge. Before an individual updates, it is necessary to determine whether there is a current desire to learn.

[0150] The formula for the primary knowledge sharing and harvesting stage is as follows:

[0151]

[0152] Denote the individual in the j-th dimension; Denote the individual in the j-th dimension, K f Denote the knowledge factor; Denote the individual in the j-th dimension, Guass(0,1) represents the Gaussian distribution with a mean of 0 and a variance of 1, Denote the j-th dimension of a random individual.

[0153] Among them, the knowledge factor K f is a real number greater than 0, controlling the total amount of knowledge obtained and shared, and this knowledge will be added to the current individual from other individuals over several generations.

[0154] Two individuals (the closest to the best and the closest to the worst) need to update a third individual. In the case of updating the best and the worst solutions, the two closest best and worst solutions are selected respectively. If x i is the global optimal solution, the following two solutions should be used to update the global optimal solution: (x i , x i+1 , x i+2 ). If x i is the global worst value, the two closest previous solutions should be used to update the global worst value: (x i , x i-1 , x i-2 ).

[0155] S3.5: The advanced knowledge sharing and harvesting stage should follow the following steps:

[0156] Sort all individuals in the current population in ascending order according to the objective function value: x1, x2, …, x NP . Among them, x1 is the solution with the smallest objective function value (the best individual), x NPThe solution with the worst objective function value (the worst individual). Then, the classified population is divided into three groups. The first group contains the best solution, the second group contains medium or better solutions, and the last group contains the worst solutions. Finally, to update each solution x1, two solutions are randomly selected from the best and worst 100p% solutions in the entire population to form the acquisition part, while the third solution is randomly selected from the middle NP - 2×100p% solutions to form the sharing part.

[0157] The formula for the advanced harvest sharing knowledge stage is as follows:

[0158]

[0159] denotes denotes the j-th dimension of the best individual of denotes the j-th dimension of the worst individual of denotes the j-th dimension of a random individual.

[0160] S3.6: After the initialization and the primary and advanced knowledge acquisition sharing stages, GSK adopts a greedy selection method to decide whether to use the old or new solution in the next iteration, which can be expressed as:

[0161]

[0162] where is the solution generated in the primary or advanced acquisition sharing knowledge stage, is the i-th solution in the current population. Among them, the selected individual is represented by denotes.

[0163] S3.7: Substitute the overall heat absorption rate of each furnace zone corresponding to each dimension of the selected individual into the two-dimensional unsteady heat transfer model of the tempering furnace in step S1 to calculate the tapping temperature of the j-th billet to be measured when tapped and the upper surface temperature of the j-th billet measured by the in-furnace pyrometer

[0164] S3.8: Then, substitute the tapping temperature of the j-th billet measured when tapped and and the upper surface temperature of the j-th billet measured by the in-furnace pyrometer into the identification objective function in step S2 to find

[0165] S3.9: After updating the global optimum using the greedy selection, determine whether the maximum number of iterations has been reached. If the maximum number of iterations has been reached, output the optimal overall heat absorption rate for each furnace zone. If the maximum number of iterations has not been reached, return to step S3.3.

[0166] In step S4, the tempering furnace operation data is used as the input, and the corresponding overall heat absorption rate is used as the output to train a recurrent neural network. The recurrent neural network is as Figure 4 shown, and the operation steps are as follows:

[0167] S4.1: Model definition. The input vector x t =[x 1,t ,x 2,t ,...,x M,t represents all relevant variables at time step t, here representing the tempering furnace operation data such as the burner duty ratio, burner fault status, fuel supply, air-fuel ratio, gas preheating temperature, steel plate speed, steel plate position, and furnace zone temperature. The size of M is determined by the type of operation data. The output vector y t represents the estimated value of the overall heat absorption rate along the furnace length at time step t. The hidden state h t ∈R m represents the internal memory of the model, where m is the number of neurons in the hidden layer.

[0168] For any given time step t, the RNN recursively calculates the new hidden state h t and the predicted value y t using the following formulas:

[0169] h t =tanh(W hx x t +W hh h t-1 +b h ) (17)

[0170] y t =W hy h t +b y (18)

[0171] Here, W hx ∈R m×n ,W hh ∈R m×m are weight matrices, and n is the number of input features. b h ∈R m and b y ∈R are the bias terms for the hidden layer and the output layer, respectively. The tanh function is used to introduce non-linearity and help the model learn more complex patterns., W hy represents the weight matrix of the output layer, x tDenotes the input layer, h t-1 Denotes the hidden layer.

[0172] S4.2: Initialize the parameters, randomly initialize all weights W hx , W hh , W hy and the bias term b h , b y . These initial values are usually drawn from a normal distribution within a small range.

[0173] S4.3: Forward propagation, given a batch of training samples (where is a series of consecutive observations, is the corresponding total hemispherical absorptivity), perform forward propagation to obtain the predicted value where, N represents the number of groups of input data, represents the output value of the T-th dimension in the training sample, represents the predicted value of the T-th dimension during the training process, and T represents the dimension of the output layer.

[0174] S4.4: Calculate the loss, select an appropriate loss function to measure the gap between the predicted value and the true value. For regression problems, common choices include mean squared error:

[0175]

[0176] S4.5: Backward propagation, use the gradient descent method to adjust the parameters to minimize the above loss function. First, the gradients with respect to W hx , W hh , W hy , b h , b y need to be calculated, and then the parameters are updated according to the selected learning rate η:

[0177]

[0178]

[0179] In the formula, represents the partial derivative of the weight, and L represents the loss function.

[0180] S4.6: Loop training, repeat the above steps for multiple rounds until the stopping condition is met (such as reaching the maximum number of iterations or the performance on the validation set no longer improves). During this period, a series of operations including forward propagation, calculating the loss, backward propagation, and updating the parameters will be performed on the entire dataset.

[0181] In step S5, the dynamically changing overall heat absorption rate output by the recurrent neural network is introduced into the heat transfer model of the tempering furnace, and a more accurate steel plate temperature prediction model is obtained.

[0182] Compared with the prior art, the steel temperature prediction model constructed by the technical solution proposed by the present invention integrates the heat transfer formula, has better interpretability, physical meaning and extrapolation than the data-driven model, and the fluctuation of the steel temperature output value is smaller. The overall heat absorption rate in the technical solution does not need to be determined in advance through coupling experiments, and the overall heat absorption rate is more in line with the actual situation than the value calculated by the empirical formula. Therefore, the model accuracy is higher than that of the mechanism model.

[0183]

[0184] Based on the above embodiments, the embodiments of the present application further provide a computer program, which when run on a computer, causes the computer to execute the method provided by the above embodiments.

[0185] Based on the above embodiments, the embodiments of the present application further provide a computer storage medium, in which a computer program is stored, and when the computer program is executed by a computer, the computer is caused to execute the method provided by the above embodiments.

[0186] Among them, the storage medium can be any available medium that can be accessed by a computer. By way of example but not limited to: the computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disc storage, magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer.

[0187] Based on the above embodiments, the embodiments of the present application further provide a chip, which is used to read the computer program stored in the memory and implement the method provided by the above embodiments.

[0188] Based on the above embodiments, the embodiments of the present application provide a computer program product, which when run on an electronic device, implements the method provided by the above embodiments.

[0189] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt 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 code.

[0190] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices produce a means for realizing the functions specified in one or more of the flows Figure 1 one or more flows and / or blocks Figure 1 or a means for realizing the functions specified in one or more of the blocks.

[0191] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including an instruction means, and the instruction means realizes the functions specified in one or more of the flows Figure 1 one or more flows and / or blocks Figure 1 or a means for realizing the functions specified in one or more of the blocks.

[0192] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for realizing the functions specified in one or more of the flows Figure 1 one or more flows and / or blocks Figure 1 or a means for realizing the functions specified in one or more of the blocks.

[0193] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these changes and modifications.

Claims

1. A method for constructing a steel temperature prediction model for the tempering process based on the full-factor drive fusion mechanism, characterized in that, including constructing the steel temperature prediction model for the tempering process using a recurrent neural network; obtaining a training data set; wherein, the input data of the steel temperature prediction model for the tempering process in the training data set is the operation data of the tempering furnace, and the output data is the total heat absorption rate; training the steel temperature prediction model for the tempering process with the training set; wherein, the training data set is obtained based on the following method: S3.

1. Initialize a population of size NP, and randomly initialize the variables of the first individual x1 within its specified range; S3.

2. Obtain the x sequence through the Chebyshev mapping as the initialized population; S3.

3. Sort all individuals in the current population in ascending order according to the objective function values: x1, x2, …, x NP , update the evaluation times F E = NP, where x1 is the individual with the optimal solution of the objective function value, and x NP is the individual with the worst solution of the objective function value; S3.

4. (1) When the random number rand is greater than the knowledge ratio K r do not update the individual, calculate the fitness value of the objective function, and update the evaluation times F E = F E + NP, and update the population; (2) When the random number rand is not greater than the knowledge ratio K r make a judgment: When the random number rand is not less than the number of dimensions N for updating the primary benefit sharing knowledge scheme junior the individual is updated using the formula in the primary knowledge sharing stage calculate the fitness value of the objective function and update the evaluation times F E =F E +NP, and update the population; (2 - 2) The random number rand is less than the number of dimensions N updated by the primary revenue - sharing knowledge scheme junior When this occurs, update the individual using the formula in the advanced knowledge - sharing stage , calculate the fitness value of the objective function, and update the evaluation times F E =F E +NP, and update the population; S3.

5. Use the greedy selection method to determine whether to use the old individual or the new individual in the next iteration; S3.

6. The selected individuals Substitute the overall heat absorption rate of each furnace zone corresponding to each dimension into the two-dimensional unsteady heat transfer model of the tempering furnace to calculate the outgoing temperature of the j-th steel plate measured when it exits the furnace and the temperature of the upper surface of the j-th steel plate measured by the high-temperature thermometer in the furnace ; S3.

7. Substitute the tapping temperature of the j-th steel plate measured at tapping and the upper surface temperature of the j-th steel plate measured by the in-furnace pyrometer and into the identification objective function to solve for the optimal overall heat absorption rate in the furnace zone; S3.

8. After updating the global optimum using the greedy selection method, determine whether the maximum number of iterations is reached. If the maximum number of iterations is reached, output the optimal total heat absorption rate of each furnace zone and the operation data of the tempering furnace for constructing the training data set; if the maximum number of iterations is not reached, return to step S3.

3.

2. The method for constructing a steel temperature prediction model for the tempering process based on the full-factor drive fusion mechanism according to claim 1, characterized in that, wherein: In step S3.1, initialize a population of size NP, and randomly initialize the variables of the first individual x1 within its specified range, as shown in the following formula: (10) where \(x\) i is the \(i\)-th individual, \(i = 1, 2, \ldots, NP\), denotes the \(j\)-th dimension of the individual \(x\) i ; \(l_j\) and \(u_j\) are the lower and upper bounds of the \(j\)-th decision variable, respectively, and \(r_j\) is a random number within the range \([0, 1]\); In step S3.2, obtain the x sequence through the Chebyshev mapping as the initialized population, as shown in the following formula: (11) wherein, is the (i + 1)-th individual, represents the order; In step S3.4, formula update in the primary knowledge sharing stage: (14) In the formula, represents the j-th dimension of the individual ; represents the j-th dimension of the individual ; represents the knowledge factor; represents the j-th dimension of the individual ; represents a Gaussian distribution with a mean of 0 and a variance of 1, represents the j-th dimension of the random individual; In step S3.4, formula update in the advanced knowledge sharing stage: (15) In the formula, represents the j-th dimension of the best individual , represents the j-th dimension of the worst individual , represents the j-th dimension of the random individual; In step S3.5, use the greedy selection method to determine whether to use the old individual or the new individual in the next iteration, as shown in the following formula: (16) In the formula, represents the selected individual, is the solution generated in the primary and advanced stages of obtaining shared knowledge, is the i-th solution in the current population.

3. The method for constructing a steel temperature prediction model for the tempering process of the all-factor driven fusion mechanism according to any one of claims 1-2, characterized in that The operation data of the tempering furnace includes any one or combination of the duty ratio of the burner, the fault state of the burner, the fuel supply and air-fuel ratio, the gas preheating temperature, the steel plate speed, the steel plate position, the furnace zone temperature, the measured length of the steel plate, the measured width, the measured thickness, the steel plate number, the steel plate type, the steel plate inlet temperature, the surface temperature of the steel plate measured by the high-temperature thermometer in the furnace, the steel plate outlet temperature, the steel plate running speed, the furnace length, the heating time, the holding time, and the temperature measured by the thermocouple in each furnace zone.

4. The method for constructing a steel temperature prediction model for the tempering process of the full-factor drive fusion mechanism according to claim 3, characterized in that The operation data of the tempering furnace includes the duty ratio of the burner, the fault state of the burner, the fuel supply and air-fuel ratio, the gas preheating temperature, the steel plate speed, the steel plate position, and the furnace zone temperature.

5. The method for constructing a steel temperature prediction model for the tempering process of the full-factor drive fusion mechanism according to any one of claims 1-2, characterized in that, The mathematical expression of the identification objective function is as follows: (8) (9) In the formula, J represents the objective function, N represents the number of steel plates, represents the tapping temperature of the j-th steel plate measured at tapping; represents the upper surface temperature of the j-th steel plate measured by the high-temperature thermometer in the furnace; represents the tapping temperature of the j-th steel plate calculated by the heat transfer model of the tempering furnace; represents the upper surface temperature of the j-th steel plate calculated by the heat transfer model of the tempering furnace when it is transported to the position where the high-temperature thermometer in the furnace is located; W represents the combined vector of the overall heat absorption rate at different positions along the furnace length, represents the m-th overall heat absorption rate along the furnace length, and λ1 and λ2 represent the regularization coefficients.

6. The method for constructing a steel temperature prediction model during the tempering process of the full-factor drive fusion mechanism according to any one of claims 1-2, characterized in that The two-dimensional unsteady heat transfer model of the tempering furnace is established based on the following method: S1.

2. Establish a two-dimensional unsteady heat conduction equation, and the specific mathematical description is as follows: (1) where ρ represents the density of the steel plate; C p represents the specific heat of the steel plate; T represents the temperature of the steel plate; τ represents time; represents the thermal conductivity of the steel plate; represents the differential along the length direction of the steel plate, represents the differential in the thickness direction of the steel plate; S1.

3. Establish boundary conditions, and the mathematical description of the heat flux density on the steel plate surface is: (2) Wherein, q represents the heat flux density on the surface of the steel plate; represents the Boltzmann constant; represents the total heat absorptivity; T f represents the furnace temperature, T s represents the temperature on the surface of the steel plate; S1.

4. According to the position of any steel plate changing with time in the furnace, the boundary conditions of the steel plate are transformed into a time-varying temperature field problem. Among them, for any steel plate i, the position x of the tail of the steel plate in the furnace i is determined by the moving speed of the steel plate and is expressed as follows: (3) In the formula, represents the position of the tail of the steel plate in the furnace, represents the time variable, t f is the residence time of the steel plate in the furnace, represents the time differential, and ν(t) is the moving speed of the steel plate at time t; S1.

5. Calculate the boundary conditions, as shown below: (4) (5) (6) (7) In the formula, represents the temperature of the steel plate, represents the position of the tail of the steel plate, represents the coordinate axis along the length direction of the steel plate, represents the coordinate axis in the thickness direction of the steel plate, represents the thickness of the steel plate, represents the length of the steel plate, q h and q t respectively represent the heat flux density on the front side and the back side of the steel plate, q u and q l respectively represent the heat flux density on the upper surface and the lower surface of the steel plate; S1.

6. Discretize the two-dimensional unsteady heat conduction equation and boundary conditions using the finite difference method, thereby iterating the heat transfer model of the two-dimensional unsteady tempering furnace and outputting the relevant temperatures at each moment. Among them, the relevant temperatures include the temperature of the j-th steel plate measured when it exits the furnace and the temperature of the upper surface of the j-th steel plate measured by the high-temperature thermometer in the furnace as well as the temperature inside the steel plate.

7. The method for constructing a steel temperature prediction model in the tempering process of the full-factor drive fusion mechanism according to any one of claims 1-2, characterized in that, The steel temperature prediction model in the tempering process is a recurrent neural network, and the input vector of the model represents the state of the tempering furnace operation data at time t, M represents the number of types of tempering furnace operation data, and the output vector represents the estimated value of the overall heat absorption rate along the furnace length direction at time t; For any given time step t, the RNN recursively computes the new hidden state via the following formula and the predicted value : (17) (18) In the formula, , represents the weight matrix, represents the weight matrix of the output layer, represents the input layer, represents the hidden layer, n represents the number of input features, and represent the bias terms for the hidden layer and the output layer respectively. The tanh function is used to introduce non-linearity, and m represents the number of neurons in the hidden layer; The model training includes S4.

1. Initialize parameters and randomly initialize all weights , , and bias terms , ; S4.

2. Forward propagation, given a batch of training samples , where is a series of consecutive observations, is the corresponding total heat absorption rate, and perform forward propagation to obtain the predicted value ; where N represents the number of groups of input data, represents the output value of the th dimension in the training sample, represents the predicted value of the th dimension during the training process, represents the dimension of the output layer; S4.

3. Calculate the loss, as shown below: (19) In the formula, represents the loss function; S4.

4. Adjust the parameters using the gradient descent method to minimize the loss function for backpropagation, including: calculating the gradients with respect to , , , , and updating the parameters according to the learning rate η, expressed as follows: (20) (21) (22) (23) (24) wherein, represents the partial derivative with respect to the weight, represents the loss function; S4.

5. Perform cyclic training until the stop condition is met.

8. A steel temperature prediction model for the tempering process with a full-factor drive fusion mechanism, constructed by any one of the methods described in claims 1-7.

9. A method for predicting the steel temperature during the tempering process of a full-factor driven fusion mechanism, characterized in that, Implement the steel temperature prediction model for the tempering process constructed by the method described in claim 8; The method includes: Input the operation data of the tempering furnace into the steel temperature prediction model for the tempering process; The steel temperature prediction model for the tempering process outputs the total heat absorption rate of each furnace zone; Based on the overall heat absorption rate of each furnace zone, the steel temperature during the tempering process is calculated through a two-dimensional unsteady heat transfer model of the tempering furnace.

Images