Method for predicting final temperature of plate blank of aluminum alloy hot rolling heating furnace

Abstract

The invention discloses a method for predicting the final temperature of an aluminum alloy hot rolling heating furnace slab. The method comprises the following steps: S1, collecting historical thermocouple burying test data of the aluminum alloy hot rolling heating furnace slab; s2, preprocessing the historical puppet burying test data; s3, constructing a deep learning network model; s4, performing training learning on the deep learning network model to obtain an aluminum alloy hot rolling heating furnace slab center end point temperature prediction model; and S5, the final temperature of the aluminum alloy plate blank center is predicted online through the aluminum alloy hot rolling heating furnace plate blank center final temperature prediction model. According to the method, the final temperature of the aluminum alloy hot rolling heating furnace plate blank can be accurately predicted, accurate control over the tapping time is achieved, the temperature hit rate is increased, energy consumption is reduced, and the production quality is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of hot rolling heating furnaces and relates to a method for predicting the end point temperature of a slab in an aluminum alloy hot rolling heating furnace. Background Art

[0002] A hot rolling furnace is a device that heats metal slabs (such as steel and aluminum billets) to a specific temperature before hot rolling. Its function is to preheat the slabs to the ideal temperature range required for rolling, thereby ensuring the quality and performance of the rolled products. During the preheating process, accurately predicting the internal endpoint temperature of the slabs in the heating furnace is crucial. If the internal endpoint of the slab does not reach the target temperature, it will lead to product dimensional deviations, poor surface quality, and uneven internal structure, thereby reducing rolling quality. Furthermore, by accurately predicting the internal endpoint temperature of the slabs, it is also possible to precisely control the heating time and the time it takes to exit the furnace, thereby reducing unnecessary energy waste, improving energy efficiency, and lowering production costs.

[0003] The surface temperature of the slab in the heating furnace can be directly obtained by installing a thermocouple at the bottom of the slab, but the center temperature needs to be obtained by digging a hole inside the slab and burying a thermocouple. Therefore, in actual production, it is impossible to obtain the changing trend of the slab center temperature in real time.

[0004] Currently, steel companies primarily use finite difference methods based on fundamental heat transfer principles to establish a one-dimensional unsteady-state heat conduction mathematical model to predict the center temperature of slabs. However, this approach is ineffective due to the complex and volatile nature of actual production environments, and the model's adaptability is insufficient. In the aluminum alloy industry, current methods for predicting the internal temperature of aluminum slabs prior to hot rolling rely on manual methods. These methods have the following drawbacks: They cannot accurately predict temperature, leading to unstable production quality; further, they result in a more conservative temperature control strategy, increasing unnecessary furnace time and energy consumption, reducing production efficiency, and increasing production costs. Summary of the Invention

[0005] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide a method for predicting the terminal temperature of aluminum alloy hot rolling heating furnace slabs. By training and learning the historical embedded couple test data of aluminum alloy hot rolling heating furnace slabs, an endpoint prediction model of the slab center is obtained, thereby achieving accurate prediction of the terminal temperature of the heating furnace center, and then achieving precise control of the furnace discharge time, thereby improving the temperature hit rate, reducing energy consumption, and improving production quality.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] The present invention provides a method for predicting the endpoint temperature of a slab in an aluminum alloy hot rolling heating furnace, comprising the following steps:

[0008] S1, collects historical embedded couple test data of aluminum alloy hot rolling furnace slab;

[0009] S2, preprocessing the historical embedded coupler test data;

[0010] S3, build a deep learning network model;

[0011] S4, deep learning network model is trained and learned to obtain the prediction model of the center endpoint temperature of the aluminum alloy hot rolling heating furnace slab;

[0012] S5, using the aluminum alloy hot rolling furnace slab center endpoint temperature prediction model to online predict the aluminum alloy slab center endpoint temperature.

[0013] Preferably, in step S1, the historical embedded couple test data of the aluminum alloy hot rolling heating furnace slab includes the size of the test slab, the composition of the test slab, the data on the change of furnace gas temperature over time, the data on the change of test slab surface temperature over time, and the endpoint temperature of the center of the test slab.

[0014] Preferably, in step S2, the preprocessing process of the historical embedded coupler test data is as follows:

[0015] The historical embedded couple test data of aluminum alloy hot rolling furnace slabs were cleaned by manual identification and statistics to remove abnormal data.

[0016] Standardize the cleaned data;

[0017] The standardized data is stored in a structured manner to obtain a structured data set.

[0018] Preferably, in step S3, a long short-term memory network is used to construct a deep learning network model for predicting the temperature of the heating furnace slab.

[0019] Preferably, the neural network of the deep learning network model consists of an LSTM neural network and a fully connected neural network;

[0020] The input layer of the LSTM neural network has two neurons, one for furnace gas temperature and the other for test slab surface temperature; the output layer of the LSTM neural network has one neuron;

[0021] The input layer of the fully connected neural network has 11 neurons, namely LSTM neural network output, slab length, slab width, slab height, slab [Al], slab [Mg], slab [Si], slab [Cr], slab [Mn], slab [Fe], and slab [Ti]. The hidden layer of the fully connected neural network has 2 layers, the first hidden layer has 4 neurons, the second hidden layer has 2 neurons, and the output layer of the hidden layer has 1 neuron, which represents the endpoint temperature of the center of the test slab.

[0022] Preferably, the specific process of step S4 is as follows:

[0023] The data obtained in step S2 is divided into a training data set and a validation data set. The training data set is trained and learned using the deep learning network model constructed in step S3, and the training effect is verified using the validation data set. The model hyperparameters are adjusted according to the verification results, and training and verification are performed again until the deviation between the predicted temperature and the actual temperature is ≤3°C. The training is stopped to obtain a prediction model for the center endpoint temperature of the aluminum alloy hot rolling heating furnace slab.

[0024] Preferably, the optimization algorithm used in the training is the stochastic gradient descent method, the loss function is the mean square error function, and the activation function is the ReLU function.

[0025] Preferably, the process of step S5 is as follows:

[0026] The aluminum alloy hot rolling furnace slab center endpoint temperature prediction model is deployed on the aluminum alloy hot rolling furnace process control model system;

[0027] The current heat slab data is obtained, the obtained current heat slab data is structured, and the end point temperature of the current slab center is predicted using the aluminum alloy hot rolling heating furnace slab center end point temperature prediction model.

[0028] The method for predicting the endpoint temperature of the aluminum alloy hot rolling furnace slab provided by the present invention has the following beneficial effects:

[0029] The present invention can effectively improve the temperature hit rate of the slab when it leaves the furnace and reduce unnecessary heating time by accurately predicting the center endpoint temperature of the slab in the aluminum alloy hot rolling heating furnace, thereby reducing energy consumption, improving production quality, and improving the automation level of the aluminum alloy hot rolling process, thereby promoting the innovation and development of the temperature prediction technology of the aluminum alloy heating furnace. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0031] Figure 1This is a flow chart of the method for predicting the endpoint temperature of a slab in an aluminum alloy hot rolling heating furnace according to the present invention;

[0032] Figure 2 This is a deep learning network diagram of the aluminum alloy hot rolling furnace slab endpoint temperature prediction method of the present invention;

[0033] Figure 3 This is a graph showing the change in furnace gas temperature over time during the first 800 minutes of heating in the current heat cycle according to an embodiment of the present invention;

[0034] Figure 4 This is a graph showing the change trend of the slab surface temperature over time in the first 800 minutes of heating in the current heat in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to better understand the above technical solution of the present invention, the technical solution of the present invention is further described below with reference to embodiments.

[0036] Combine Figure 1 As shown, the present invention provides a method for predicting the endpoint temperature of an aluminum alloy hot rolling furnace slab, comprising the following steps:

[0037] S1, collects historical embedded couple test data of aluminum alloy hot rolling furnace slab;

[0038] This step primarily involves collecting historical embedded couple test data. This data for aluminum alloy hot-rolling furnace slabs includes test slab dimensions, test slab composition, furnace gas temperature over time, test slab surface temperature over time, and the endpoint temperature at the center of the test slab. The test slab dimensions include length, width, and height. The test slab composition includes: [Al], [Mg], [Si], [Cr], [Mn], [Fe], and [Ti].

[0039] S2, preprocessing of historical buried couplet test data;

[0040] Firstly, manual recognition and statistics are used to clean the historical embedded couple test data of aluminum alloy hot rolling furnace slabs to remove abnormal data;

[0041] The cleaned data is then standardized;

[0042] In a specific embodiment, the standardization processing method is as follows: x std represents the standardized data, x represents the sample data, μ represents the mean value of the sample data, and σ represents the standard deviation of the sample data.

[0043] Finally, the standardized data is stored in a structured manner to obtain a structured data set.

[0044] S3, build a deep learning network model;

[0045] A long short-term memory network is used to construct a deep learning network model for heating furnace slab temperature prediction.

[0046] Combine Figure 2 As shown, the neural network of the deep learning network model consists of an LSTM neural network and a fully connected neural network. The input layer of the LSTM neural network has two neurons, representing the furnace gas temperature and the test slab surface temperature, respectively; the output layer of the LSTM neural network has one neuron. The input layer of the fully connected neural network has 11 neurons, representing the LSTM neural network output, slab length, slab width, slab height, slab [Al], slab [Mg], slab [Si], slab [Cr], slab [Mn], slab [Fe], and slab [Ti]. The fully connected neural network has two hidden layers: the first hidden layer has four neurons, the second hidden layer has two neurons, and the output layer has one neuron, representing the endpoint temperature at the center of the test slab.

[0047] S4, deep learning network model is trained and learned to obtain the prediction model of the center endpoint temperature of the aluminum alloy hot rolling heating furnace slab;

[0048] The data obtained in step S2 is divided into a training data set and a validation data set. The training data set is trained and learned using the deep learning network model constructed in step S3, and the training effect is verified using the validation data set. The model hyperparameters are adjusted according to the verification results, and training and verification are performed again until the deviation between the predicted temperature and the actual temperature is ≤3°C. The training is stopped to obtain the center endpoint temperature prediction model of the aluminum alloy hot rolling heating furnace slab.

[0049] The optimization algorithm used in the above training is the stochastic gradient descent method, the loss function is the mean square error function, and the activation function is the ReLU function.

[0050] S5, using the aluminum alloy hot rolling furnace slab center endpoint temperature prediction model to online predict the aluminum alloy slab center endpoint temperature.

[0051] The aluminum alloy hot rolling furnace slab center endpoint temperature prediction model is deployed on the aluminum alloy hot rolling furnace process control model system; then online prediction is performed: first, the current furnace slab data is obtained, then the obtained current furnace slab data is structured, and finally the aluminum alloy hot rolling furnace slab center endpoint temperature prediction model is used to predict the current slab center endpoint temperature.

[0052] The method for predicting the end point temperature of the aluminum alloy hot rolling furnace slab of the present invention will be further introduced below with reference to specific examples.

[0053] Example

[0054] This example was carried out in a domestic aluminum alloy processing plant. The method of the present invention was used to predict the center endpoint temperature of the slab in the hot rolling heating furnace. The specific process is as follows:

[0055] (1) Collect historical embedded couple test data: including embedded couple test slab dimensions: slab length, slab width, slab height; content of each alloy element in the slab material: slab [Al], slab [Mg], slab [Si], slab [Cr], slab [Mn], slab [Fe], slab [Ti]; data on the change of furnace gas temperature over time; data on the change of slab surface temperature over time, and the end point temperature of the slab center.

[0056] (2) Preprocessing of historical couplet test data

[0057] The data obtained in step 1 is preprocessed. First, abnormal data is removed by manual screening. Then the filtered data is standardized using the following method:

[0058]

[0059] Among them, x std represents the standardized data, x represents the sample data, μ represents the mean value of the sample data, and σ represents the standard deviation of the sample data. Finally, the filtered data is structured and stored to obtain a structured data set.

[0060] (3) Building a deep learning network model

[0061] Construct a deep learning network model, whose structure is as follows Figure 2 As shown in the figure, the neural network consists of an LSTM neural network and a fully connected neural network. The LSTM neural network input layer has two neurons, representing the furnace gas temperature and the slab surface temperature, respectively, and the LSTM neural network output layer has one neuron. The fully connected neural network input layer has 11 neurons, representing the LSTM neural network output, slab length, slab width, slab height, slab [Al], slab [Mg], slab [Si], slab [Cr], slab [Mn], slab [Fe], and slab [Ti]. There are two hidden layers, the first hidden layer has four neurons, the second hidden layer has two neurons, and the output layer has one neuron, representing the slab center endpoint temperature.

[0062] (4) Deep learning model training

[0063] The data obtained in step 2 was divided into a training dataset and a validation dataset. The deep learning network model constructed in step 3 was trained on the training dataset, and the training results were verified using the validation dataset. Based on the validation results, the model hyperparameters were adjusted, and training and validation were repeated. When the results showed that the deviation between the model's predicted temperature and the actual temperature was 3°C, and the validation results met the expected requirements, training was stopped, and a model for predicting the center endpoint temperature of aluminum alloy hot-rolling furnace slabs was obtained. The training used the stochastic gradient descent optimization algorithm, the mean squared error loss function, and the ReLU activation function.

[0064] (5) Online prediction of slab temperature

[0065] The aluminum alloy hot rolling heating furnace slab center endpoint temperature prediction model is deployed on the aluminum alloy hot rolling heating furnace process control model system; then the aluminum alloy hot rolling heating furnace slab center endpoint temperature prediction model is used to predict the hot rolling heating furnace slab center endpoint temperature, as follows:

[0066] Get the current heat slab data, see Table 1;

[0067] Table 1

[0068]

[0069] The trend of furnace gas temperature over time in the first 800 minutes of slab heating in the current heat is as follows: Figure 3 As shown in the figure, the surface temperature of the slab in the first 800 minutes of heating of the current heat changes with time. Figure 4 As shown;

[0070] Using the above data as input conditions, the aluminum alloy hot rolling furnace slab center endpoint temperature prediction model is used to predict the slab center temperature after heating for 800 minutes. The prediction result is 399.2℃.

[0071] This example used this method to predict the center endpoint temperature of slabs in a hot-rolling furnace at a domestic aluminum alloy processing plant. The predictions were validated using an embedded couple test. The results showed a 3°C deviation between the model's predicted and actual temperatures. As shown in Table 2, the qualified rate of hot-rolled export slabs increased by 0.9% three months after the model was implemented, and the average furnace time was reduced by 5%, significantly improving production efficiency and quality while reducing production costs.

[0072] Table 2

[0073] Pass rate (%) Time in furnace (min) Before the model goes online 98.5 873 After the model is launched 99.4 829

[0074] Those skilled in the art should recognize that the above embodiments are merely intended to illustrate the present invention and are not intended to limit the present invention. As long as they are within the spirit of the present invention, any changes or modifications to the above embodiments will fall within the scope of the claims of the present invention.

Claims

1. A method for predicting the end point temperature of a slab in an aluminum alloy hot rolling heating furnace, characterized in that: The following steps are involved: S1, collect historical buried couple test data of aluminum alloy hot rolling heating furnace slab; S2, preprocessing the historical embedded couple test data; S3, build a deep learning network model; S4, the deep learning network model is trained and learned to obtain a prediction model for the center endpoint temperature of the slab in the aluminum alloy hot rolling heating furnace; S5, using the aluminum alloy hot rolling heating furnace slab center end temperature prediction model to online predict the aluminum alloy slab center end temperature.

2. The method for predicting the end point temperature of the aluminum alloy hot rolling heating furnace slab according to claim 1, characterized in that: In step S1, the historical embedded couple test data of the aluminum alloy hot rolling heating furnace slab includes the size of the test slab, the composition of the test slab, the data on the change of furnace gas temperature over time, the data on the change of test slab surface temperature over time, and the endpoint temperature of the center of the test slab.

3. The method for predicting the end point temperature of the aluminum alloy hot rolling heating furnace slab according to claim 1, characterized in that: In step S2, the preprocessing process of the historical embedded couple test data is as follows: The historical embedded couple test data of aluminum alloy hot rolling heating furnace slabs were cleaned by manual identification and statistics to remove abnormal data. Standardize the cleaned data; The standardized data is stored in a structured manner to obtain a structured data set.

4. The method for predicting the end point temperature of the aluminum alloy hot rolling heating furnace slab according to claim 1, characterized in that: In step S3, a long short-term memory network is used to construct a deep learning network model for predicting the temperature of a heating furnace slab.

5. The method for predicting the end point temperature of the aluminum alloy hot rolling heating furnace slab according to claim 4, characterized in that: The neural network of the deep learning network model consists of an LSTM neural network and a fully connected neural network; The input layer of the LSTM neural network has two neurons, which are the furnace gas temperature and the test slab surface temperature; the output layer of the LSTM neural network has one neuron; The input layer of the fully connected neural network has 11 neurons, which are LSTM neural network output, slab length, slab width, slab height, slab [Al], slab [Mg], slab [Si], slab [Cr], slab [Mn], slab [Fe], and slab [Ti]; the hidden layer of the fully connected neural network has 2 layers, the first hidden layer has 4 neurons, the second hidden layer has 2 neurons, and the output layer of the hidden layer has 1 neuron, which represents the endpoint temperature of the center of the test slab.

6. The method for predicting the end point temperature of the aluminum alloy hot rolling heating furnace slab according to claim 1, characterized in that: The specific process of step S4 is as follows: The data obtained in step S2 is divided into a training data set and a verification data set. The training data set is trained and learned using the deep learning network model constructed in step S3, and the training effect is verified using the verification data set. The model hyperparameters are adjusted according to the verification results, and training and verification are performed again until the deviation between the predicted temperature and the actual temperature is ≤3°C. The training is stopped to obtain a prediction model for the center endpoint temperature of the aluminum alloy hot rolling heating furnace slab.

7. The method for predicting the end point temperature of the aluminum alloy hot rolling heating furnace slab according to claim 6, characterized in that: The optimization algorithm used in the training is the stochastic gradient descent method, the loss function is the mean square error function, and the activation function is the Relu function.

8. The method for predicting the end point temperature of the aluminum alloy hot rolling heating furnace slab according to claim 1, characterized in that: The process of step S5 is as follows: The prediction model of the slab center endpoint temperature of the aluminum alloy hot rolling heating furnace is deployed on the process control model system of the aluminum alloy hot rolling heating furnace; The slab data of the current furnace is obtained, and the obtained slab data of the current furnace is structured and processed, and the endpoint temperature of the center of the slab of the aluminum alloy hot rolling heating furnace is predicted using the endpoint temperature prediction model of the center of the slab.

Images