Tension prediction model construction method, tension prediction method and device, and medium
The tension prediction model constructed through the deep neuron network solves the problem of inaccurate tension control of traditional cold rolling flattening machines, realizes precise control of different steel types, and improves production efficiency and product quality.
Patent Information
- Application Number
- CN202411035970.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-08
AI Technical Summary
Traditional cold rolling flattening machines have insufficient precise control in tension control, and cannot effectively identify the performance differences between different steel types, resulting in limited production efficiency and product quality.
The tension prediction model is constructed by deep neuron networks. By extracting the yield strength, roughness, elongation and thickness of the rolled material as feature parameter data, the feature parameter data set is constructed, and the pre-constructed tension neuron network model is trained. Evaluate and optimize, and obtain the optimized tension prediction model.
It improves the intelligence level of the cold rolling flattening machine production process, can automatically extract useful information from a large amount of data, adapt to different steel types and performance requirements, and improves product quality and production efficiency.
Smart Images

Figure CN118904936A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of rolling technology, and in particular to a tension prediction model construction method, a tension prediction method and device, and a medium. Background Art
[0002] In the field of cold rolling and skin-pass mill technology, tension control is a key factor to ensure stable strip rolling and obtain high-quality products. In the rolling process of traditional cold rolling and skin-pass mill, the tension is mainly controlled by the speed difference between the stands. This control method ensures the stability of the rolling process to a certain extent, but due to the lack of accurate control algorithms and models, the tension control effect is subject to certain limitations, which in turn affects the quality and production efficiency of the product.
[0003] For products of different steel grades, although they may belong to the same type (such as DX51D), their performance and control targets vary greatly. Traditional control schemes often cannot effectively identify these differences, resulting in the inability to achieve precise control during the rolling process. For example, the yield strength of DX51D products may span about 100Mpa, but traditional systems cannot distinguish these subtle differences, making it difficult to accurately control according to the actual performance requirements of the product.
[0004] In addition, the pre-calculation of control parameters of the traditional skin-pass mill rolling process is mainly based on skin-pass mill rolling theory, but these theoretical models often ignore the various influencing factors and product influencing coefficients in actual working conditions, resulting in low adaptability and intelligence of the model. And continuous optimization and control can only be carried out manually, which not only increases the workload, but also limits the improvement of production efficiency and product quality. Summary of the invention
[0005] In view of the above problems, the present invention is proposed to provide a tension prediction model construction method, tension prediction method and device, and medium that overcome the above problems or at least partially solve the above problems. By constructing a tension prediction model, the model is based on a deep neural network, which can widely identify control requirements and apply intelligent technology to solve existing model self-optimization and adaptability problems, thereby improving the intelligence level of the cold rolling mill production process and significantly improving product quality and production efficiency.
[0006] In a first aspect, a tension prediction model construction method is provided for predicting the tension of a leveling machine, the method comprising:
[0007] Acquire historical rolling source data of the skin-pass mill, extract characteristic parameter data affecting tension and tension corresponding to the characteristic parameter data from the historical rolling source data, wherein the characteristic parameter data at least includes yield strength, roughness, elongation and thickness of the rolled material;
[0008] Constructing a characteristic parameter data set according to the characteristic parameter data and the corresponding tension;
[0009] The pre-constructed tension neural network model is trained, evaluated and optimized according to the characteristic parameter data set to obtain an optimized tension prediction model.
[0010] Optionally, constructing a characteristic parameter data set according to the characteristic parameter data and the corresponding tension includes:
[0011] The characteristic parameter data is used as input data, and the tension corresponding to the characteristic parameter is used as target output data, so as to construct a characteristic parameter data set including the input data and the target output data.
[0012] Optionally, the pre-built tension neural network model is a deep neural network model, comprising four input layers, a hidden layer and an output layer, wherein the input features of the input layer are the feature parameter data, and the output result of the output layer is the predicted tension.
[0013] Optionally, the pre-built tension neural network model is trained, evaluated and optimized according to the characteristic parameter data set to obtain an optimized tension prediction model, including:
[0014] Dividing the characteristic parameter data set into a training set and a test set;
[0015] Based on the training set, training the pre-built tension neural network model;
[0016] Based on the test set, evaluating the stability and prediction accuracy of the trained tension neural network model to obtain an evaluation result;
[0017] When the evaluation result shows that the tension neural network model does not meet the standard, the tension neural network model is optimized according to a set optimization strategy to obtain an optimized tension prediction model.
[0018] Optionally, based on the test set, evaluating the prediction accuracy of the trained tension neural network model includes:
[0019] Inputting the characteristic parameter data in the test set into the trained tension neural network model to obtain the predicted tension output by the tension neural network model;
[0020] Determine a difference between the predicted tension and a target tension, wherein the target tension is the tension corresponding to the input characteristic parameter data in the test set;
[0021] When the difference is within a preset difference threshold range, outputting the tension neural network model as meeting the standard;
[0022] When the difference is outside the preset difference threshold range, it is output that the tension neural network model does not meet the standard.
[0023] Optionally, when the evaluation result is that the tension prediction model does not meet the standard, optimizing the tension prediction model according to a set optimization strategy includes:
[0024] When the evaluation result is that the tension prediction model does not meet the standard, determining that the type of rolled product corresponding to the characteristic parameter data input into the tension prediction model is a target type product;
[0025] Acquire historical rolling source data when the skin-pass mill rolls the target product, extract characteristic parameter data and corresponding tension corresponding to the target product from the historical rolling source data, and construct an optimization set;
[0026] Based on the optimization set, the tension neural network model is optimized by using an incremental self-learning method.
[0027] In a second aspect, a tension prediction model construction device is provided for predicting the tension of a leveling machine, the device comprising:
[0028] A parameter extraction module, used to obtain historical rolling source data of the skin-pass mill, and extract characteristic parameter data affecting the tension and the tension corresponding to the characteristic parameter data from the historical rolling source data, wherein the characteristic parameter data at least includes the yield strength, roughness, elongation and thickness of the rolled material;
[0029] A data set building module, used to build a feature parameter data set according to the feature parameter data and the corresponding tension;
[0030] The model building module is used to train, evaluate and optimize the pre-built tension neural network model according to the characteristic parameter data set to obtain an optimized tension prediction model.
[0031] In a third aspect, a tension prediction method is provided for predicting the tension of a leveling machine, the method comprising:
[0032] Extracting target characteristic parameter data from the material properties of the product to be rolled, wherein the target characteristic parameter data includes yield strength, roughness, elongation and thickness of the material to be rolled;
[0033] The target characteristic parameters are input into the tension prediction model constructed according to the method described in the first aspect to obtain predicted tension.
[0034] In a fourth aspect, a tension prediction device is provided for predicting the tension of a leveling machine, the device comprising:
[0035] A parameter extraction module, used to extract target characteristic parameter data from the material characteristics of the product to be rolled, wherein the target characteristic parameter data includes yield strength, roughness, elongation and thickness of the material to be rolled;
[0036] A tension prediction module is used to input the target characteristic parameters into the tension prediction model constructed according to the method described in the first aspect to obtain predicted tension.
[0037] In a fifth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the tension prediction model construction method as described in the first aspect.
[0038] The technical solution provided in the embodiments of the present invention has at least the following technical effects or advantages:
[0039] The embodiments of the present invention provide a tension prediction model construction method, tension prediction method and device, and medium. The method extracts the yield strength, roughness, elongation, and thickness of the rolled material as characteristic parameter data affecting the tension, and constructs a characteristic parameter data set together with the tension corresponding to the characteristic parameter data. Then, the pre-constructed tension neural network model is trained, evaluated, and optimized according to the constructed characteristic parameter data set to obtain an optimized tension prediction model for predicting the tension of the leveling machine. The model is based on a deep neural network, which can widely identify control requirements and apply intelligent technology to solve existing model self-optimization and adaptability problems. Through the learning ability of the deep neural network, the model can automatically extract useful information from a large amount of production and experimental data, and continuously optimize its own parameters to adapt to products with different steel grades and performance requirements. At the same time, the model also has self-learning and self-adaptation capabilities, and can use the latest production data for self-optimization and improvement, improve the accuracy and reliability of the model, which is conducive to improving product quality and production efficiency.
[0040] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:
[0042] Figure 1It is a flow chart of a tension prediction model construction method provided by an embodiment of the present invention;
[0043] Figure 2 is a structural schematic diagram of a neural network model provided by an embodiment of the present invention;
[0044] Figure 3 is a schematic diagram of a loss function used in a model training process provided by an embodiment of the present invention;
[0045] Figure 4 is a schematic diagram of the difference between a predicted tension and a target tension provided by an embodiment of the present invention;
[0046] Figure 5 is a flow chart of a model incremental self-learning process provided by an embodiment of the present invention;
[0047] Figure 6 is a structural block diagram of a tension prediction model building device provided by an embodiment of the present invention;
[0048] Figure 7 is a flow chart of a tension prediction method provided by an embodiment of the present invention;
[0049] Figure 8 It is a structural block diagram of a tension prediction device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0050] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings.
[0051] Various structural schematic diagrams according to embodiments of the present disclosure are shown in the accompanying drawings. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clear expression. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may further design regions / layers with different shapes, sizes, and relative positions according to actual needs.
[0052] In order to better understand the above-mentioned technical scheme, the above-mentioned technical scheme will be described in detail below in combination with specific implementation methods. It should be understood that the embodiments of the present disclosure and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.
[0053] Figure 1 is a flow chart of a tension prediction model construction method provided by an embodiment of the present invention. Figure 1As shown, the method includes steps S110 to S130, which are specifically as follows:
[0054] Step S110, obtaining historical rolling source data of the leveling mill, extracting characteristic parameter data affecting tension and tension corresponding to the characteristic parameter data from the historical rolling source data, wherein the characteristic parameter data at least includes yield strength, roughness, elongation and thickness of the rolled material.
[0055] Specifically, historical rolling source data within a set number or a set time period can be obtained according to different product types. For example, for Class A products, the first X1 source data of historical rolling of the same Class A product, or all source data within the historical time period t1 (t can be several months or several years, etc., which is not limited to this) can be obtained, and each source data includes the above-mentioned characteristic parameter data and the tension corresponding to the characteristic parameter data; for Class B products, the first X2 source data of historical rolling of the same Class B product, or all source data within the historical time period t2 can be obtained. Specifically, it can be obtained according to actual needs, and this embodiment does not limit this.
[0056] In this embodiment, considering the same type of steel, the performance and control targets may vary greatly. For example, the yield strength of the most common DX51D product spans about 100Mpa. Each specific product is different, but they are all DX51D in the system. In the process of implementing the present invention, it is found that the differentiated requirements for product performance are mainly related to the parameters of yield strength, roughness, elongation and thickness of the rolled material. Therefore, the present invention extracts the above parameters as characteristic parameter data to construct a tension prediction model to meet the differentiated control requirements of product performance as much as possible and improve the adaptability and budget accuracy of the model.
[0057] Optionally, before executing step S120, the method may further include:
[0058] Preprocess the extracted characteristic parameter data.
[0059] Specifically, this can include removing noise, outliers, missing values, etc. from the data to ensure the quality and consistency of the data. Perform necessary transformations on the data, such as scaling and normalization, so that the model can better process the data.
[0060] Step S120: construct a characteristic parameter data set according to the characteristic parameter data and the corresponding tension.
[0061] Specifically, the characteristic parameter data is used as input data, and the tension corresponding to the characteristic parameter is used as target output data, so as to construct a characteristic parameter data set including the input data and the target output data.
[0062] Step S130: training, evaluating and optimizing the pre-built tension neural network model according to the characteristic parameter data set to obtain an optimized tension prediction model.
[0063] Optional, pre-built tensor neural network model for deep neural network model, Figure 2 is a schematic diagram of the structure of a neural network model provided by an embodiment of the present invention, such as Figure 2 As shown in the figure, the deep neural network model includes an input layer (x1~xn), a hidden layer (h11~h1n, ..., hm1~hmn), and an output layer (y1~yn). The input layer is the beginning of the neural network, which receives raw data or features. The number of neurons in the input layer usually matches the number of features of the input data. There can be one or more hidden layers between the input layer and the output layer. The number of hidden layers and the number of neurons in each layer can be adjusted according to the complexity of the problem and the required learning ability. The layers are fully connected, that is, each neuron in each layer is connected to each neuron in the next layer. In each hidden layer, the neurons perform a weighted sum of the inputs and apply an activation function (such as ReLU, Sigmoid, or tanh) to produce outputs. These outputs will serve as inputs to the next layer.
[0064] In this embodiment, the pre-built tension neural network model includes four input layers, a hidden layer and an output layer, wherein the input features of the input layer are feature parameter data, and the output result of the output layer is the predicted tension. By using the deep neural network model to build a tension budget model, accurate prediction and intelligent control of tension can be achieved.
[0065] Optionally, step S130 includes:
[0066] The first step is to divide the feature parameter data set into a training set and a test set;
[0067] The second step is to train the pre-built tension neural network model based on the training set;
[0068] In this embodiment, the trained tension neural network model can be stored to facilitate subsequent evaluation and optimization of the model.
[0069] The third step is to evaluate the stability and prediction accuracy of the trained tension neural network model based on the test set and obtain the evaluation results;
[0070] Since the training process of the model is to find the minimum value of the loss function. Deep learning frameworks (such as TensorFlow, PyTorch, etc.) are usually used to train the model. Training requires adjusting hyperparameters such as network structure, optimizer, and learning rate to achieve the performance requirements of the optimized model and make it better adapt to the actual operating environment of the cold rolling mill. At the same time, in order to prevent overfitting and improve the generalization ability of the model, some regularization techniques such as L1 regularization and L2 regularization can be used. The parameters of the model are continuously adjusted through the training set to gradually reduce the loss function. Therefore, the stability of the trained tension neural network model can be evaluated according to the loss function. Among them, the loss function is a function used to measure the difference between the model prediction value and the true value. Figure 3 is a schematic diagram of a loss function used in a model training process provided by an embodiment of the present invention, such as Figure 3 As shown in the figure, after selecting a suitable loss function, during the model training process, each round of iteration will calculate a loss value based on the current model parameters. This loss value is a specific quantification of the difference between the model prediction value and the true value. During the training process, it is necessary to constantly monitor the change of the loss value. If the loss value gradually decreases, it means that the model is gradually learning and optimizing; if the loss value fluctuates or increases, the model parameters may need to be adjusted. In addition to monitoring the change of the loss value on the training set, the test set is also needed to evaluate the performance of the model. Calculating the loss value on the test set can more objectively reflect the generalization ability of the model. If the loss value of the model on the test set also gradually decreases, it means that the model has good generalization ability; if the loss value on the test set increases, it may be overfitting, and measures need to be taken to prevent overfitting. In order to more accurately evaluate the performance of the model, the loss value of the current model can be compared with the loss value of the benchmark model. If the loss value of the current model is significantly lower than the loss value of the benchmark model, it means that the current model has better stability.
[0071] Optionally, based on the test set, the prediction accuracy of the trained tensor neural network model is evaluated, including:
[0072] The characteristic parameter data in the test set is input into the trained tension neural network model to obtain the predicted tension output by the tension neural network model; the difference between the predicted tension and the target tension is determined, and the target tension is the tension corresponding to the input characteristic parameter data in the test set; when the difference is within the preset difference threshold range, the output tension neural network model meets the standard; when the difference is outside the preset difference threshold range, the output tension neural network model does not meet the standard.
[0073] Figure 4 is a schematic diagram of the difference between a predicted tension and a target tension provided by an embodiment of the present invention, such as Figure 4 As shown, the horizontal axis represents the predicted tension F预 , the vertical axis represents the target tension F 目 , when the point is located on the line S, it means that the difference between the predicted tension and the target tension is 0. The preset difference threshold range can be set according to actual needs and is not limited thereto.
[0074] It should be noted that in this embodiment, the model is evaluated mainly by the two indicators of loss function and prediction accuracy. In practice, other evaluation indicators may be combined, such as recall rate, F1 value, etc., which are not limited in the present invention. When the evaluation result of the model from any indicator is not up to standard, the output evaluation result is not up to standard.
[0075] Step 4: When the evaluation result shows that the tension neural network model does not meet the standards, the tension neural network model is optimized according to the set optimization strategy to obtain an optimized tension prediction model.
[0076] On the contrary, when the evaluation result shows that the tension neural network model meets the standard, there is no need to optimize the tension neural network model, and it can be directly used as the tension prediction model.
[0077] In an implementation of this embodiment, when the evaluation result is that the tension prediction model does not meet the standard, the tension prediction model is optimized according to the set optimization strategy, including:
[0078] When the evaluation result shows that the tension prediction model does not meet the standards, the type of rolled product corresponding to the characteristic parameter data input into the tension prediction model is determined to be the target product; the historical rolling source data when the leveler rolls the target product is obtained, and the characteristic parameter data and the corresponding tension corresponding to the target product are extracted from the historical rolling source data to construct an optimization set; based on the optimization set, the tension neural network model is optimized using an incremental self-learning method.
[0079] By optimizing the model using incremental self-learning, the model's ability to quickly learn new tasks can be enhanced, making the model able to adapt to changes in real time. When new data flows in, the model can be updated and optimized immediately to quickly adapt to changing data. Incremental self-learning can continue to learn new data based on the original model without retraining the entire model, thereby improving the generalization effect of the model and reducing the risk of overfitting. At the same time, the incremental self-learning model needs to store less data at a time, which helps save memory and storage space and achieve efficient use of resources.
[0080] It should be noted that the newly introduced data of the incremental self-learning model, that is, the data in the optimization set, can also be quality controlled and preprocessed to avoid the impact of noise data and outliers on the model. At the same time, online learning techniques such as small batch updates and learning rate decay can also be used to improve model accuracy. Ultimately, the model can adapt to changes in product characteristics, thereby improving the model's prediction accuracy and generalization ability.
[0081] Figure 5 is a flow chart of a model incremental self-learning process provided by an embodiment of the present invention, such as Figure 5 As shown, after receiving the steel coil output telegram, the Java class is loaded by executing an external process to execute the executable exe file. The executable exe file executes the online self-learning process, that is, by connecting to the Oracle database to calculate the deviation between the predicted value and the actual value, and write it to the database, and then read the historical data of similar products to evaluate the model accuracy to determine whether the model meets the standard. When the model meets the standard, the database link is ended and no incremental self-learning is performed; when the model does not meet the standard, the category (i.e., the target product) is marked and written to the database for data collection. If the incremental data collected is greater than or equal to 1000 and the production line is stopped, the model is incrementally self-learned, and then the latest model accuracy is evaluated to meet the standard. After meeting the standard, the data table is cleared, the database link is ended, and the incremental self-learning is completed; if the incremental data is not greater than or equal to 1000, or the production line is not stopped, the database link is directly ended without incremental self-learning.
[0082] In other implementations of this embodiment, the tension prediction model may be optimized by adjusting hyperparameters such as the network structure, optimizer, and learning rate so that the model can adapt to different product requirements and thus maintain its good prediction performance.
[0083] In this embodiment, the optimized model can be embedded into the existing production line process control database and integrated into the production line process control system to achieve real-time tension budgeting.
[0084] The embodiment of the present invention innovatively applies the deep neural network (DNN) model structure to the pre-calculation of the tension of the cold rolling mill, describes its complex nonlinear relationship, and effectively improves the fitting ability and generalization ability of the model. And the established deep neural network model can be embedded in any existing control system through a friendly user interface to achieve seamless integration with the existing production process. Intelligent regulation of the production process of the cold rolling mill is achieved by real-time monitoring and adjustment of the tension value. According to the deviation situation, the real-time data of products with large deviations is dynamically collected to prepare for the next step of dynamic optimization and adjustment of the model. At the same time, it can also provide strong support for the production decision-making of the enterprise, help the enterprise to achieve refined and intelligent production management, and improve the core competitiveness of the enterprise.
[0085] Based on the same inventive concept, an embodiment of the present invention provides a device for implementing the tension prediction model construction method in the above embodiment. Figure 6 is a structural block diagram of a tension prediction model building device provided by an embodiment of the present invention, such as Figure 6 As shown, the device 600 includes a parameter extraction module 610 , a data set establishment module 620 and a model establishment module 630 .
[0086] The parameter extraction module 610 is used to obtain the historical rolling source data of the leveling mill, and extract the characteristic parameter data affecting the tension and the tension corresponding to the characteristic parameter data from the historical rolling source data, wherein the characteristic parameter data at least includes the yield strength, roughness, elongation and thickness of the rolled material. The data set establishment module 620 is used to construct a characteristic parameter data set based on the characteristic parameter data and the corresponding tension. The model establishment module 630 is used to train, evaluate and optimize the pre-constructed tension neural network model based on the characteristic parameter data set to obtain an optimized tension prediction model.
[0087] Optionally, the data set establishing module 620 is used to:
[0088] The characteristic parameter data is taken as input data, the tension corresponding to the characteristic parameter is taken as target output data, and a characteristic parameter data set including the input data and the target output data is constructed.
[0089] Optionally, the pre-built tension neural network model is a deep neural network model, including four input layers, a hidden layer and an output layer, wherein the input features of the input layer are feature parameter data, and the output result of the output layer is the predicted tension.
[0090] Optionally, the model building module 630 includes:
[0091] A data set division unit, used for dividing the feature parameter data set into a training set and a test set;
[0092] A model training unit, used to train a pre-built tensor neural network model based on a training set;
[0093] A model evaluation unit is used to evaluate the stability and prediction accuracy of the trained tensor neural network model based on the test set to obtain an evaluation result;
[0094] The model optimization unit is used to optimize the tension neural network model according to the set optimization strategy to obtain an optimized tension prediction model when the evaluation result shows that the tension neural network model does not meet the standards.
[0095] Optionally, the model evaluation unit is also used to:
[0096] The characteristic parameter data in the test set is input into the trained tension neural network model to obtain the predicted tension output by the tension neural network model; the difference between the predicted tension and the target tension is determined, and the target tension is the tension corresponding to the input characteristic parameter data in the test set; when the difference is within the preset difference threshold range, the output tension neural network model meets the standard; when the difference is outside the preset difference threshold range, the output tension neural network model does not meet the standard.
[0097] Optionally, the model optimization unit is also used to:
[0098] When the evaluation result shows that the tension prediction model does not meet the standards, the type of rolled product corresponding to the characteristic parameter data input into the tension prediction model is determined to be the target product; the historical rolling source data when the leveler rolls the target product is obtained, and the characteristic parameter data and the corresponding tension corresponding to the target product are extracted from the historical rolling source data to construct an optimization set; based on the optimization set, the tension neural network model is optimized using an incremental self-learning method.
[0099] The specific details of the tension prediction model construction method used in the above-mentioned device can be understood by corresponding to the relevant descriptions and effects in the above-mentioned tension prediction model construction method embodiment, and will not be repeated here.
[0100] Based on the same inventive concept, an embodiment of the present invention provides a tension prediction method, which uses the tension prediction model constructed in the above embodiment to predict the tension of a leveling machine. Figure 7 is a flow chart of a tension prediction method provided by an embodiment of the present invention, such as Figure 7 As shown, the method includes:
[0101] Step S710: extract target characteristic parameter data from the material properties of the product to be rolled, where the target characteristic parameter data includes the yield strength, roughness, elongation and thickness of the material to be rolled.
[0102] Step S720: input the target characteristic parameters into a pre-built tension prediction model to obtain predicted tension.
[0103] The tension prediction model pre-built in step S720 is composed of Figure 1 The model has been applied to the tension prediction of the leveler of Shougang Jingtang United Iron and Steel Co., Ltd. 1-5# galvanizing, 8# galvanizing, 1# continuous annealing, and 2# continuous annealing in the secondary process control system in the form of a database link.
[0104] Based on the same inventive concept, an embodiment of the present invention further provides a device for implementing the tension prediction method in the above embodiment. Figure 8 is a structural block diagram of a tension prediction device provided by an embodiment of the present invention, such as Figure 8 As shown, the device 800 includes a parameter extraction module 810 and a tension prediction module 820 .
[0105] The parameter extraction module 810 is used to extract target characteristic parameter data from the material characteristics of the product to be rolled, and the target characteristic parameter data includes the yield strength, roughness, elongation and thickness of the material to be rolled. The tension prediction module 820 is used to input the target characteristic parameters into a pre-built tension prediction model to obtain the predicted tension.
[0106] Among them, the tension prediction model pre-constructed in this embodiment is composed of Figure 1 The specific details of the tension prediction model construction method can be understood by corresponding to the relevant descriptions and effects in the tension prediction model construction method embodiment shown above, and will not be repeated here.
[0107] Optionally, the present invention also provides a computer product, which includes a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute to implement the tension prediction model construction method of the above method embodiment.
[0108] The technical solutions provided in the above embodiments of the present application have at least the following technical effects or advantages:
[0109] The embodiments of the present invention provide a tension prediction model construction method, tension prediction method and device, and medium. The method extracts the yield strength, roughness, elongation, and thickness of the rolled material as characteristic parameter data affecting the tension, and constructs a characteristic parameter data set together with the tension corresponding to the characteristic parameter data. Then, the pre-constructed tension neural network model is trained, evaluated, and optimized according to the constructed characteristic parameter data set to obtain an optimized tension prediction model for predicting the tension of the leveling machine. The model is based on a deep neural network, which can widely identify control requirements and apply intelligent technology to solve existing model self-optimization and adaptability problems. Through the learning ability of the deep neural network, the model can automatically extract useful information from a large amount of production and experimental data, and continuously optimize its own parameters to adapt to products with different steel grades and performance requirements. At the same time, the model also has self-learning and self-adaptation capabilities, and can use the latest production data for self-optimization and improvement, improve the accuracy and reliability of the model, which is conducive to improving product quality and production efficiency.
[0110] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.
[0111] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those explicitly recited in each claim. More specifically, as reflected in the claims below, inventive aspects lie in less than all the features of the individual embodiments disclosed above. Therefore, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, with each claim itself serving as a separate embodiment of the present invention.
[0112] It should be noted that the above-mentioned embodiments illustrate rather than limit the invention and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims.
Claims
1. A tension prediction model construction method, characterized in that: For predicting the tension of a skin-pass mill, the method comprises: Acquire historical rolling source data of the skin-pass mill, extract characteristic parameter data affecting tension and tension corresponding to the characteristic parameter data from the historical rolling source data, wherein the characteristic parameter data at least includes yield strength, roughness, elongation and thickness of the rolled material; Constructing a characteristic parameter data set according to the characteristic parameter data and the corresponding tension; The pre-constructed tension neural network model is trained, evaluated and optimized according to the characteristic parameter data set to obtain an optimized tension prediction model.
2. The method according to claim 1, characterized in that: The step of constructing a characteristic parameter data set according to the characteristic parameter data and the corresponding tension includes: The characteristic parameter data is used as input data, and the tension corresponding to the characteristic parameter is used as target output data, so as to construct a characteristic parameter data set including the input data and the target output data.
3. The method according to claim 1, characterized in that The pre-constructed tension neural network model is a deep neural network model, including four input layers, a hidden layer and an output layer, wherein the input feature of the input layer is the feature parameter data, and the output result of the output layer is the predicted tension.
4. The method according to claim 1, characterized in that: The pre-built tension neural network model is trained, evaluated and optimized according to the characteristic parameter data set to obtain an optimized tension prediction model, including: Dividing the characteristic parameter data set into a training set and a test set; Based on the training set, training the pre-built tension neural network model; Based on the test set, evaluating the stability and prediction accuracy of the trained tension neural network model to obtain an evaluation result; When the evaluation result shows that the tension neural network model does not meet the standard, the tension neural network model is optimized according to a set optimization strategy to obtain an optimized tension prediction model.
5. The method according to claim 4, characterized in that Based on the test set, the prediction accuracy of the trained tension neural network model is evaluated, including: Inputting the characteristic parameter data in the test set into the trained tension neural network model to obtain the predicted tension output by the tension neural network model; Determine a difference between the predicted tension and a target tension, wherein the target tension is the tension corresponding to the input characteristic parameter data in the test set; When the difference is within a preset difference threshold range, outputting the tension neural network model as meeting the standard; When the difference is outside the preset difference threshold range, it is output that the tension neural network model does not meet the standard.
6. The method according to claim 5, characterized in that When the evaluation result shows that the tension prediction model does not meet the standard, optimizing the tension prediction model according to a set optimization strategy includes: When the evaluation result is that the tension prediction model does not meet the standard, determining that the type of rolled product corresponding to the characteristic parameter data input into the tension prediction model is a target type product; Acquire historical rolling source data when the skin-pass mill rolls the target product, extract characteristic parameter data and corresponding tension corresponding to the target product from the historical rolling source data, and construct an optimization set; Based on the optimization set, the tension neural network model is optimized by using an incremental self-learning method.
7. A tension prediction model construction device, characterized in that: For predicting the tension of a skin-pass mill, the device comprises: A parameter extraction module, used to obtain historical rolling source data of the skin-pass mill, and extract characteristic parameter data affecting the tension and the tension corresponding to the characteristic parameter data from the historical rolling source data, wherein the characteristic parameter data at least includes the yield strength, roughness, elongation and thickness of the rolled material; A data set building module, used to build a characteristic parameter data set according to the characteristic parameter data and the corresponding tension; The model building module is used to train, evaluate and optimize the pre-built tension neural network model according to the characteristic parameter data set to obtain an optimized tension prediction model.
8. A tension prediction method, characterized in that: For predicting the tension of a skin-pass mill, the method comprises: Extracting target characteristic parameter data from the material properties of the product to be rolled, wherein the target characteristic parameter data includes yield strength, roughness, elongation and thickness of the material to be rolled; The target characteristic parameters are input into the tension prediction model constructed according to the method according to any one of claims 1 to 6 to obtain predicted tension.
9. A tension prediction device, characterized in that: For predicting the tension of a skin-pass mill, the device comprises: A parameter extraction module, used to extract target characteristic parameter data from the material characteristics of the product to be rolled, wherein the target characteristic parameter data includes yield strength, roughness, elongation and thickness of the material to be rolled; A tension prediction module is used to input the target characteristic parameters into the tension prediction model constructed according to the method described in any one of claims 1 to 6 to obtain predicted tension.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the tension prediction model construction method according to any one of claims 1 to 6.
Citation Information
Patent Citations
Cold rolling force prediction method and system based on machine learning method
CN111889524A
Intelligent tension control method and system for metal rolling mill
CN115846421A
Casting blank defect online prediction method
CN116822315A
Rod and wire rolling force prediction method and device based on finite element and neural network
CN117473833A
Cited By
Steam bath high-power drafting parameter dynamic prediction system and method
CN120524756A
Rolling schedule adaptive calculation method and system based on neural network
CN120562478A
PID (Proportion Integration Differentiation)-based crude foil engine multi-stage tension cooperative control method and system
CN120742658A
