Method and device for predicting convexity of strip steel of finishing mill group in hot rolling production process in real time

Abstract

The invention discloses a real-time prediction method and device for the convexity of strip steel of a finishing mill group in the hot rolling production process, and belongs to the technical field of ferrous metallurgy. The method comprises the steps that multiple pieces of mechanical variable data influencing the convexity of the strip steel of the finishing mill group in the hot rolling production process are obtained; performing sliding window reconstruction on the obtained mechanical variable data; constructing a sample data set by using the mechanical variable data reconstructed by the sliding window; a convolution gating circulation unit is improved, and the improved convolution gating circulation unit is used for constructing a strip steel convexity prediction model of the finishing mill group in the hot rolling production process; training the constructed strip steel convexity prediction model of the finishing mill group in the hot rolling production process; and utilizing the trained strip steel convexity prediction model of the finishing mill group in the hot rolling production process to realize strip steel convexity prediction. According to the scheme, the technical problem that the real-time performance and accuracy of hot rolling strip steel finishing mill group plate convexity prediction cannot be achieved in the prior art can be effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of iron and steel metallurgy, and particularly to a method and device for real-time prediction of strip convexity in a finishing mill unit during the hot rolling production process. Background Art

[0002] The hot rolling production line mainly consists of a heating furnace, a high-pressure water descaling device, a width sizing press, a roughing mill, a finishing mill, a laminar cooling device, a coiler, etc. Hot rolled strip is a steel product made by rolling high-temperature steel billets. During the rolling process, the thickness, width, and shape of the sheet may change. Therefore, predicting the shape of the final product is crucial for meeting quality standards. In the iron and steel industry, especially in the production of hot rolled strip, shape prediction is a crucial task. The above complex process flow determines that the strip convexity is affected by many production factors, such as the geometric and material properties of the metal itself, rolling speed, bending roll force, bending roll form, roll wear, etc. In the work of plate convexity prediction, problems such as numerous and highly coupled parameters, and extensive time-varying non-linear effects in hot rolled strip have always been important factors affecting the prediction accuracy. Shape prediction is one of the key technologies in the fields of industrial manufacturing and material processing, aiming to predict the shape and characteristics of the final product sheet during the manufacturing process. This prediction process is crucial for ensuring product quality, production efficiency, and resource utilization. Shape prediction usually relies on mathematical modeling, data analysis, and previous production data. By considering factors such as process parameters, mechanical properties, temperature, and pressure, the prediction model can help production managers predict the sheet characteristics of the final product, so as to timely adjust the production process to meet quality standards. Through shape prediction, production enterprises can identify potential problems in advance, avoid production defects, reduce the scrap rate, improve production efficiency, and ensure product consistency. This not only helps to save costs but also enhances the competitiveness of the enterprise. Developing a core model for a data-driven hot rolling plate convexity control system, formulating reasonable and effective plate convexity control strategies, and constructing an intelligent plate convexity control system applicable to hot rolled strip production are of great significance for improving the control level of domestic hot continuous rolling strip.

[0003] Common hot-rolled strip crown prediction methods can be divided into mechanism analysis-based methods and data-driven prediction methods. Usually, an on-line strip crown prediction model is established based on a mechanism analysis model. The mechanism analysis model first constructs a preliminary model according to the physical laws of the hot-rolling process (including temperature field distribution, stress-strain analysis, dynamic simulation, etc.), and then realizes the prediction of the crown based on the physical modeling results combined with specific process data. The data-driven method usually uses machine learning methods to construct a prediction model. To avoid the problems of high data dependence and easy overfitting, many researchers have improved machine learning algorithms and carried out prediction modeling based on the improved machine learning algorithms. In recent years, with the rapid development of artificial intelligence and its successful application in various fields such as industry and commerce, more and more scholars and technical experts have introduced data-driven methods into the rolling prediction model and achieved remarkable results. By analyzing the production process data and using statistical principles, the action rules between production variables are mined to obtain a high-precision prediction model that meets expectations. At the same time, the rapid development of intelligent sensing devices, intelligent control systems, and computer storage and computing capabilities also provides a technical basis for the application of data-driven methods in complex process industrial systems such as hot rolling. However, the mechanism model only considers the single influence of various influencing factors on the strip crown at the exit under basic process parameters, but these factors are not independent of each other, so the prediction accuracy of the traditional model is not high; most of the commonly used data-driven models at the present stage focus on global feature extraction or time series feature extraction, and are relatively lacking in the ability to extract local features and spatial features, and the prediction accuracy is also relatively low.

[0004] Therefore, the existing technology has the technical problems of being unable to realize the real-time prediction of the strip crown of the finishing mill unit in the hot-rolled strip production process and the low prediction accuracy. Therefore, it is necessary to improve the existing prediction method for the strip crown of the finishing mill unit in the hot-rolled strip production process. Summary of the Invention

[0005] The present invention provides a real-time prediction method and device for the strip crown of the finishing mill unit in the hot-rolling production process to solve the technical problems that the existing technology cannot realize the real-time prediction of the strip crown of the finishing mill unit in the hot-rolled strip production process and the low prediction accuracy.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] On the one hand, the present invention provides a real-time prediction method for the strip crown of the finishing mill unit in the hot-rolling production process, and the real-time prediction method for the strip crown of the finishing mill unit in the hot-rolling production process includes:

[0008] Obtain multiple mechanical variable data that affect the strip crown of the finishing mill unit in the hot-rolling production process;

[0009] Based on the sliding window operation, the obtained mechanical variable data is reconstructed by sliding window;

[0010] Using the mechanical variable data reconstructed by sliding window to construct a sample data set;

[0011] Improve the convolutional gated recurrent unit, and use the improved convolutional gated recurrent unit to construct a strip crown prediction model for the finishing mill unit in the hot rolling production process;

[0012] Use the sample data set to train the strip crown prediction model for the finishing mill unit in the hot rolling production process; wherein, the input of the strip crown prediction model for the finishing mill unit in the hot rolling production process is the mechanical variable data reconstructed by sliding window, and the output is the strip crown of the finishing mill unit in the hot rolling production process;

[0013] Use the trained strip crown prediction model for the finishing mill unit in the hot rolling production process to realize strip crown prediction.

[0014] Furthermore, the obtaining of multiple mechanical variable data that affect the strip crown of the finishing mill unit in the hot rolling production process includes:

[0015] Use sensors in the on-site hot rolling strip finishing mill production line to collect actual production data;

[0016] According to the correlation between variables and crown, screen the collected actual production data, and select a preset number of mechanical property variable data with the highest correlation with the strip crown of the finishing mill unit in the hot rolling production process.

[0017] Furthermore, the reconstructing of the obtained mechanical variable data by sliding window based on the sliding window operation includes:

[0018] Represent the mechanical variable data as P*N-dimensional data with the values of each mechanical variable as the vertical axis and the data time series as the horizontal axis; where N represents the number of mechanical variables and P represents the length of the time series;

[0019] Perform a sliding window operation on the vertical axis data at each time step to reconstruct the data at each time step into multiple subsequences composed of M features; where M is the sliding window length.

[0020] Furthermore, the improvement of the convolutional gated recurrent unit includes:

[0021] Combine the convolutional block attention module with the convolutional layer in the convolutional gated recurrent unit.

[0022] Further, when the improved convolutional gated recurrent unit performs feature extraction, first, the convolutional layer replaces the fully connected layer in the convolutional gated recurrent unit to enhance the local feature extraction ability, and then the convolutional block attention module enhances the expressive ability of the convolutional layer to further improve the local feature extraction ability of the model.

[0023] Further, after using the trained hot rolling process finishing mill strip crown prediction model to realize strip crown prediction, the real-time hot rolling process finishing mill strip crown prediction method further includes:

[0024] Evaluating the model prediction ability using a preset evaluation index.

[0025] Further, the preset evaluation index is one or a combination of more than one of root mean square error RMSE, mean absolute error MAE, mean relative error MRE, and coefficient of determination R2.

[0026] On the other hand, the present invention also provides a real-time hot rolling process finishing mill strip crown prediction device, and the real-time hot rolling process finishing mill strip crown prediction device includes:

[0027] A data processing module for:

[0028] Obtaining a plurality of mechanical variable data that affect the strip crown of the hot rolling process finishing mill;

[0029] Based on a sliding window operation, performing sliding window reconstruction on the obtained mechanical variable data;

[0030] Using the mechanical variable data after sliding window reconstruction to construct a sample data set;

[0031] A model construction module for improving the convolutional gated recurrent unit and using the improved convolutional gated recurrent unit to construct a hot rolling process finishing mill strip crown prediction model;

[0032] A model training module for training the hot rolling process finishing mill strip crown prediction model constructed by the model construction module using the sample data set constructed by the data processing module;

[0033] A real-time hot rolling process finishing mill strip crown prediction module for realizing strip crown prediction using the hot rolling process finishing mill strip crown prediction model trained by the model training module.

[0034] On yet another aspect, the present invention also provides an electronic device, which includes a processor and a memory; wherein, at least one instruction is stored in the memory, and the instruction is loaded and executed by the processor to implement the above method.

[0035] In another aspect, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored, and the instruction is loaded and executed by a processor to implement the above method.

[0036] The beneficial effects brought by the technical solution provided by the present invention at least include:

[0037] 1. Based on the sliding window method, the present invention reconstructs the feature dimension of the convexity data of hot-rolled strip steel, making it easier to extract the local features and the local correlation between variables;

[0038] 2. The present invention incorporates a convolutional neural network into a recurrent neural network, enabling the prediction model to extract both the local features and the temporal features of the data simultaneously;

[0039] 3. The present invention combines the convolutional block attention module with the convolutional gated recurrent unit, enhancing the expressive ability of the convolutional layer in the network, enhancing the ability to extract local features, and improving the prediction accuracy;

[0040] 4. The hot-rolled strip steel finishing mill strip convexity prediction model established by the present invention based on the improved convolutional gated recurrent unit combines the advantages of the gated recurrent unit and the convolutional neural network, can better capture the temporal features and local features of the data in the billet finishing process, not only enhances the real-time performance, but also improves the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0042] Figure 1 is a schematic execution flow diagram of the strip convexity real-time prediction method for the finishing mill in the hot-rolling production process provided by the embodiment of the present invention;

[0043] Figure 2 is a process flow diagram of the finishing mill in the hot-rolled strip steel production process;

[0044] Figure 3 is a geometric definition diagram of the strip convexity of the hot-rolled strip steel finishing mill;

[0045] Figure 4 is a partial variable list diagram provided by the embodiment of the present invention;

[0046] Figure 5 is a schematic diagram of the sliding window operation provided by the embodiment of the present invention;

[0047] Figure 6This is a schematic diagram of the convolutional block attention module structure;

[0048] Figure 7 is a schematic diagram of the structure of a convolutional gated recurrent unit provided by an embodiment of the present invention;

[0049] Figure 8 Schematic diagram of the system structure of a real-time prediction device for the convexity of a strip of a finishing mill in a hot rolling production process provided by an embodiment of the present invention;

[0050] Figure 9 It is a system block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0052] First of all, it should be noted that in the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "exemplarily" is intended to present the concept in a concrete way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0053] First embodiment

[0054] This embodiment provides a method for real-time prediction of the convexity of the strip steel of the finishing mill group in the hot rolling production process, so as to achieve real-time prediction of the convexity of the strip steel plate of the rolling mill group in the hot rolling strip production process, and solve the problem that the existing technology cannot achieve the real-time and accurate prediction of the convexity of the strip steel of the finishing mill group in the hot rolling strip production process. The method can be implemented by an electronic device, which can be a terminal or a server. Its execution process is as follows Figure 1 As shown, the following steps are included:

[0055] S1, obtaining multiple mechanical variable data that affect the convexity of the strip steel of the finishing mill in the hot rolling production process;

[0056] Specifically, in this embodiment, the above S1 implementation process is as follows:

[0057] S11, using sensors in the on-site hot strip finishing mill production line to collect actual production data of the finishing mill during the hot strip production process, including roll gap, stand rolling force, rolling speed, etc.;

[0058] Among them, the production process of hot strip finishing mill is as follows: Figure 2 As shown; the plate convexity reflects the geometric shape of the hot-rolled strip, and its definition is as follows Figure 3As shown, for the cross-section of the strip perpendicular to the rolling direction after hot rolling, 90% of the middle section has the characteristics of an approximate quadratic curve. There is an edge thinning phenomenon in the hot rolling process of the strip, which will cause the thickness of the strip to decrease sharply when approaching the edge of the hot-rolled strip. Therefore, when generally measuring or calculating the plate crown, the area affected by edge thinning is usually ignored. When actually calculating the plate crown, generally a reference point is selected at each position on both sides of the strip close to the edge thinning area, and the thickness at this point represents the thickness of the strip edge. In industrial production and scientific research, the reference point is usually 25 mm or 40 mm away from the edge. The calculation formula for the plate crown can be expressed as (the reference point is 40 mm away from the edge):

[0059]

[0060] In the formula, h c is the thickness at the center, h i and h i ′ are the thicknesses at the reference points on both sides respectively, and the unit is millimeter.

[0061] S12, variable screening is performed on the original data according to the correlation between variables and crown, and 108 mechanical property variables with the highest correlation with crown are selected. Some of the variables are as Figure 4 shown.

[0062] S2, based on the sliding window operation, the obtained mechanical variable data is reconstructed by the sliding window;

[0063] It should be noted that different from the ordinary sliding window that reconstructs the time series dimension, the sliding window operation in this embodiment reconstructs the feature dimension. The specific operation is as Figure 5 shown. The left side is the original input data, where the vertical axis represents the mechanical property characteristics related to the crown of the hot-rolled strip, and the horizontal axis represents the data time series. The input data dimension is P*N. Among them, N represents the number of selected features, and P represents the length of the time series. Figure 5 The original data in shows the data with the dimension of P*N; the dotted box in the data after the sliding window processing represents the sliding window acting on one time step. Let the length of the sliding window be M and the sliding step be S. Then, after performing the sliding window operation on the feature dimension for the data at each time step, the feature sequence data dimension corresponding to each time step T P is:

[0064]

[0065] That is, the dimension of the finally obtained reconstructed input data is:

[0066]

[0067] Among them, represents rounding down.

[0068] Similar to the sliding window operation in the time series dimension, after performing the sliding window operation in the feature dimension to reconstruct the data, the data at each time step will be reconstructed into multiple subsequences composed of M features. This processing method can enhance the model's ability to sense the local correlation of the features of the hot-rolled strip process data, thereby further improving the model's expression ability in local feature extraction; at the same time, it can also enhance the diversity of data features and improve the generalization ability of the model.

[0069] S3. Construct a sample data set using the mechanical variable data reconstructed by the sliding window.

[0070] S4. Improve the convolutional gated recurrent unit, and use the improved convolutional gated recurrent unit to construct a prediction model for the strip crown of the finishing mill in the hot-rolling production process.

[0071] Specifically, the implementation process of the prediction model for the strip crown of the finishing mill in the hot-rolling production process is as follows:

[0072] S41. Encode the mechanical property characteristics and the measured crown values of the billet based on the recurrent neural network constructed by the convolutional gated recurrent unit; specifically as follows:

[0073] S411. Use the crown data processed by the sliding window as the input variable of the backbone network, with the size and shape of where N represents the number of selected features, P represents the length of the time series, and M represents the length of the sliding window;

[0074] S412. The convolutional gated recurrent unit performs the fusion of spatio-temporal features. The specific fusion method is that the convolutional part processes the spatial structure in the input data, and the GRU part processes the time dependence, and the fusion of spatio-temporal features is achieved through the combination of convolutional operations and recurrent units.

[0075] S413. Output the result after spatio-temporal feature fusion through the output layer to obtain a network prediction result with the size and shape of (P,1). Its meaning is that the predicted strip crown of the finishing mill of the hot-rolled strip is obtained through spatio-temporal feature fusion, that is, for each billet of the finishing mill of the hot-rolled strip, there is the crown of the billet corresponding to different measurement points.

[0076] S42. Improve the convolutional gated recurrent unit based on the convolutional block attention module to achieve the extraction of local features of the data; specifically as follows:

[0077] This embodiment mainly focuses on improving the ability of the prediction model to extract the spatial and local features of the convexity data of hot-rolled strip steel. In the convolutional gated recurrent unit, the extraction of spatial and local features is mainly completed by the convolutional layer therein. As an attention module for strengthening the convolutional layer, the convolutional block attention module can effectively improve the expressive ability of the convolutional layer in the network and enhance the ability of the network to extract spatial and local features. To enhance the convolutional layer in the model specifically, this embodiment proposes an improved convolutional gated recurrent unit, which combines the convolutional block attention module with the convolutional layer in the original ConvGRU, thereby improving the ability of the convolutional layer in the network to extract spatial and local features.

[0078] The method to maximize the local feature extraction ability of the proposed model is as follows: First, the fully connected layer in the traditional gated recurrent unit is replaced by a convolutional layer to enhance the local feature extraction ability, and then the convolutional block attention module enhances the expressive ability of the convolutional layer to further improve the local feature extraction ability of the model.

[0079] By introducing channel attention and spatial attention mechanisms, CBAM can suppress the less important features in the data, making the convolutional layer more focused on the key features, thereby enhancing the expressive ability of the convolutional layer. This enhancement enables the network to not only extract more discriminative features but also better adapt to complex tasks, improving the performance and generalization ability of the model. The structure of CBAM is as Figure 6 shown.

[0080] S421, Input the feature map F into CBAM. Its number of channels is C, and the height and width are H and W respectively, which can be expressed as:

[0081] F∈R C×H×W

[0082] S422, After receiving the input, CBAM processes the input using the channel attention module and the spatial attention module respectively;

[0083] The focus of the channel attention module is on the relationship between features in each channel, mainly used to extract the meaningful content information in the input and compress the spatial information in the input. In CBAM, the channel attention module uses both max pooling and average pooling at the same time. The result of average pooling is F avg , reflecting the global information of the input; the result of max pooling is F max , reflecting the prominent features of the input. Then these two feature maps are fed into a shared network composed of a multi-layer perceptron (MLP) to calculate the channel attention map M C . The calculation formula is as follows:

[0084] M C = σ(MLP(AvgPool(F)) + MLP(MaxPool(F)))

[0085] = σ(W1(W0(F avg )) + W1(W0(F max )))

[0086] In the formula, σ represents the sigmoid function, MLP is the multi-layer perceptron, AvgPool and MaxPool represent average pooling and max pooling respectively, and W0 and W1 are parameters in the multi-layer perceptron model.

[0087] S424. The calculation process of the spatial attention module is similar to that of the channel attention module. Its focus is on the spatial relationship of features and is mainly used to extract the position information of the target in the input, complementing the channel attention module. In CBAM, the spatial attention module also needs to perform max pooling and average pooling to obtain two two-dimensional feature maps, then splice the two obtained feature maps and perform a convolution operation to generate the spatial attention map M S , thus highlighting the target area. The calculation formula is as follows:

[0088]

[0089] In the formula, f represents the splicing and convolution operations.

[0090] S425. Sum the results of its two attention modules and connect them to obtain the required feature information M:

[0091] M = M C + M S

[0092] S43. Based on the cyclic structure of the convolutional gated recurrent unit, realize the extraction of the temporal features of the data;

[0093] Among them, it should be noted that ConvGRU is a recurrent neural network unit that combines the characteristics of both convolutional neural networks and gated recurrent units. Its unit structure is as Figure 7 shown. GRU is a variant of the recurrent neural network (RNN) and is mainly used to process temporal data. It can effectively capture the long-term dependencies in sequential data by introducing a gating mechanism, while avoiding the problem of gradient vanishing or explosion in traditional RNNs. The gating mechanism of GRU (update gate and reset gate) allows the network to selectively remember or forget information when processing sequences. Through recursive processing over time, GRU can effectively extract the dependencies between different time steps, thus capturing the trend of the strip crown changing over time during the production process.

[0094] The calculation process of each ConvGRU unit is as follows:

[0095]

[0096] where h t-1 is the hidden state of the previous unit; is the activation value of the new hidden state; h t is the hidden state of the current unit; x t is the input of the current unit; z t is the activation value of the update gate, and the closer the value of z t is to 1, the more information from the previous moment is retained; r t is the activation value of the reset gate, and the closer r t is to 0, the more information from the previous moment needs to be forgotten; * represents the convolution operation; b z , b r , b h are hyperparameters that ConvGRU needs to learn, namely the convolution kernel weights of the input data, the convolution kernel weights of the hidden state, and the bias.

[0097] S44, based on the convolutional gated recurrent unit, combines the convolutional layer and the recurrent structure for spatio-temporal feature fusion of the extracted temporal features and spatial features;

[0098] It should be noted that the advantage of ConvGRU lies in its spatio-temporal feature fusion ability. ConvGRU realizes spatio-temporal feature fusion through the combination of convolution operations and recurrent units. The convolutional part can handle the spatial structure in the input data, while the GRU part handles the time dependence. The combination of the two enables ConvGRU to not only effectively extract the temporal features of the plate crown data but also take into account the spatial features and local features of the data.

[0099] ConvGRU applies convolutional operations to the calculation of the update and candidate states of GRU, rather than simply performing one-dimensional convolutional processing on the original data. This enables each GRU cell to capture not only temporal information but also local information through convolution; at each time step, ConvGRU processes the state from the previous moment and the local features of the current input (obtained through convolution). These information are passed and updated through the gating mechanism of the GRU cell (i.e., the update gate and the reset gate), so as to be able to predict the future convexity changes in the temporal dimension while retaining the details in the local feature dimension. At each moment, the convolutional layer of ConvGRU extracts local features by performing convolutional operations on the input strip convexity data. At each moment, the convolutional layer of ConvGRU extracts local features by performing convolutional operations on the input strip convexity data.

[0100] After the convolutional layer extracts local features, these local features are fed into the input end of the GRU cell to calculate the update and candidate states. Through this mechanism, ConvGRU can combine the local feature information of convolution with the temporal dynamic information of GRU, thereby effectively learning and predicting the patterns of spatio-temporal changes and achieving the fusion of spatio-temporal features. Through such spatio-temporal feature fusion, ConvGRU can more accurately predict the convexity changes of the strip during the hot rolling process, effectively avoiding the problem that traditional models cannot handle spatial and temporal information simultaneously.

[0101] S5, using the sample data set to train the strip convexity prediction model of the finishing mill unit in the hot rolling production process; wherein, the input of the strip convexity prediction model of the finishing mill unit in the hot rolling production process is the mechanical variable data reconstructed by the sliding window, and the output is the strip convexity of the finishing mill unit in the hot rolling production process;

[0102] S6, using the trained strip convexity prediction model of the finishing mill unit in the hot rolling production process to achieve strip convexity prediction;

[0103] S7, using the preset evaluation index to evaluate the model prediction ability;

[0104] Specifically, in this embodiment, the implementation process of the above S7 is as follows:

[0105] The root mean squared error (RMSE), mean absolute error (MAE), mean relative error (MRE), and coefficient of determination (R-square, R2) are selected as the evaluation indicators for the model prediction results. RMSE, MAE, and MRE are used to measure the prediction accuracy. The smaller the values of these three, the higher the prediction accuracy of the model; R2 is used to reflect the quality of the model, and its value range is [0,1]. The closer R2 is to 1, the better the performance of the prediction model. The calculation formulas are as follows:

[0106]

[0107]

[0108] In the formula, y i represents the true value of the i-th sample; represents the predicted value of the i-th sample; represents the average value of the true values of the samples; n is the total number of samples.

[0109] In summary, this embodiment provides a real-time prediction method for strip crown of the finishing mill group in the hot rolling production process. The feature dimensions of the strip crown data of the hot rolled strip are reconstructed based on the sliding window method; a convolutional neural network is incorporated into the recurrent neural network, and the convolutional block attention module is combined with the convolutional gated recurrent unit, so as to combine the advantages of the gated recurrent unit and the convolutional neural network to better capture the temporal features and local features of the billet finishing process data, which not only improves the prediction real-time performance but also improves the prediction accuracy. This can not only improve the product quality but also provide operation guidance for production process personnel, and has great practical significance.

[0110] Second Embodiment

[0111] This embodiment provides a real-time prediction device for strip crown of the finishing mill group in the hot rolling production process. The system structure of the real-time prediction device for strip crown of the finishing mill group in the hot rolling production process is as Figure 8 shown, and includes the following modules:

[0112] The data processing module is used for:

[0113] Obtain multiple mechanical variable data that affect the strip crown of the finishing mill group in the hot rolling production process;

[0114] Based on the sliding window operation, perform sliding window reconstruction on the obtained mechanical variable data;

[0115] Use the mechanical variable data after sliding window reconstruction to construct a sample data set;

[0116] A model construction module, which is used to improve the convolutional gated recurrent unit and construct a strip crown prediction model for the finishing mill unit in the hot rolling production process by using the improved convolutional gated recurrent unit;

[0117] A model training module, which is used to train the strip crown prediction model for the finishing mill unit in the hot rolling production process constructed by the model construction module by using the sample data set constructed by the data processing module;

[0118] A real-time strip crown prediction module for the finishing mill unit in the hot rolling production process, which is used to realize strip crown prediction by using the strip crown prediction model for the finishing mill unit in the hot rolling production process trained by the model training module;

[0119] A model evaluation module, which is used to evaluate the model prediction ability by using preset evaluation indexes.

[0120] It should be noted that, for the sake of convenience of description, Figure 8 only the main components of the device are shown. Moreover, the real-time strip crown prediction device for the finishing mill unit in the hot rolling production process of this embodiment corresponds to the real-time strip crown prediction method for the finishing mill unit in the hot rolling production process of the first embodiment; among them, the functions realized by each functional module in the real-time strip crown prediction device for the finishing mill unit in the hot rolling production process correspond one by one to each process step in the above real-time strip crown prediction method for the finishing mill unit in the hot rolling production process; therefore, it will not be elaborated here.

[0121] The third embodiment

[0122] This embodiment provides an electronic device, as Figure 9 shown, the electronic device includes: a processor and a memory; wherein, the processor and the memory can be connected through a communication bus; at least one instruction is stored in the memory, and the instruction is loaded and executed by the processor to implement the method of the first embodiment. In addition, the electronic device may further include a transceiver, the processor and the transceiver can be connected through a communication bus, and the transceiver is used for communicating with other devices.

[0123] Next, in combination with Figure 9 specific introductions will be made to the various components of the electronic device:

[0124] Among them, the processor is the control center of the electronic device. The electronic device may include multiple processors, and each of these processors may be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The processor here may be a single processor or a collective term for multiple processing elements. For example, the processor is one or more central processing units (CPUs), or it may be other general-purpose processors, application specific integrated circuits (ASICs), or one or more integrated circuits configured to implement the embodiments of the present invention. For example: one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor can execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0125] In a specific implementation, as an embodiment, the processor may include one or more CPUs. For example Figure 9 CPU0 and CPU1 shown in [figure reference], of course, this is only an exemplary illustration.

[0126] The memory is used to store the software program for implementing the solution of the present invention and is controlled by the processor for execution. The specific implementation method may refer to the above method embodiments and will not be elaborated here.

[0127] Optionally, the memory may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, or may be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, 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, but is not limited thereto. The memory may be integrated with the processor or exist independently and be coupled to the processor through the interface circuit ( Figure 9 not shown) of the electronic device. The embodiments of the present invention do not make specific limitations thereto.

[0128] The transceiver may include a receiver and a transmitter ( Figure 9 not shown separately). Among them, the receiver is used to implement the receiving function, and the transmitter is used to implement the sending function. The transceiver may be integrated with the processor or exist independently and be coupled to the processor through the interface circuit ( Figure 9 not shown) of the electronic device. The embodiments of the present invention do not make specific limitations thereto.

[0129] In addition, it should be noted that Figure 9 the structure of the electronic device shown in does not constitute a limitation on the device. The actual device may include more or fewer components than shown in the figure, or combine some components, or have a different component layout. In addition, the technical effects achieved by the electronic device when executing the method of the first embodiment above may refer to the technical effects described in the first embodiment above, so they will not be repeated here.

[0130] Fourth Embodiment

[0131] This embodiment provides a computer-readable storage medium, in which at least one instruction is stored. The instruction is loaded and executed by the processor to implement the method of the first embodiment above. Among them, the computer-readable storage medium may be ROM, random access memory, CD-ROM, magnetic tape, floppy disk, and optical data storage devices, etc. The instructions stored therein can be loaded and executed by the processor in the terminal to implement the above method.

[0132] In addition, it should be noted that the present invention can be provided as a method, apparatus, or computer program product. Therefore, embodiments of the present invention can take the form of all or part of a hardware embodiment, all or part of a software embodiment, or an embodiment combining software and hardware aspects. Moreover, when implemented using software, embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in accordance with the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another, for example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center containing one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0133] Embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes 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, an embedded processor, or other programmable data processing terminal device to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing terminal device generate a device for implementing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0134] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1the functions specified in one or more boxes. These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in one or more processes and / or boxes. Figure 1 one process or more processes and / or boxes Figure 1 steps for implementing the functions specified in one box or more boxes.

[0135] It should also be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. The term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the existence of additional identical elements in the process, method, article or terminal device comprising the element. In addition, the term "and / or" is only a description of the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. Among them, A and B can be singular or plural. In addition, the character " / " in this article generally represents an "or" relationship between the preceding and following associated objects, but it may also represent an "and / or" relationship, which can be understood specifically with reference to the context. "At least one" means one or more, and "a plurality" means two or more. "At least one of the following (items)" or similar expressions refer to any combination of these items, including any combination of single (item) or plural items (items). For example, at least one of a, b or c can mean: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, c can be single or multiple.

[0136] In addition, it can be understood that in various embodiments of the present invention, the magnitudes of the serial numbers of the above processes do not mean the sequence of execution. The execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0137] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0138] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of functional modules / units is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of devices or units can be in electrical, mechanical, or other forms. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in each embodiment of the present invention, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0139] If the method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0140] Finally, it should be noted that the above are only the preferred embodiments of the present invention. It should be pointed out that although the preferred embodiments of the present invention have been described, for those of ordinary skill in the art, once the basic creative concept of the present invention is known, several improvements and refinements can be made without departing from the principles described in the present invention. These improvements and refinements should also be regarded as the protection scope of the present invention. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present invention.

Claims

1. A real-time prediction method for the convexity of a strip steel in a finishing mill during hot rolling production, characterized in that: include: Acquire multiple mechanical variable data that affect the strip crown of the finishing mill in the hot rolling production process; Based on the sliding window operation, the acquired mechanical variable data is reconstructed by sliding window; Construct a sample data set using the mechanical variable data reconstructed by the sliding window; The convolution gated recurrent unit is improved, and the strip crown prediction model of the finishing mill in the hot rolling production process is constructed using the improved convolution gated recurrent unit. The constructed strip convexity prediction model of the finishing mill in the hot rolling production process is trained using the sample data set; wherein the input of the strip convexity prediction model of the finishing mill in the hot rolling production process is the mechanical variable data reconstructed by the sliding window, and the output is the strip convexity of the finishing mill in the hot rolling production process; The strip crown prediction is realized by using the trained strip crown prediction model of the finishing mill in the hot rolling production process.

2. The method for real-time prediction of strip convexity of a finishing mill group in a hot rolling production process according to claim 1, characterized in that: The acquisition of multiple mechanical variable data that affect the strip convexity of the finishing mill group in the hot rolling production process includes: Use sensors in the on-site hot strip finishing mill production line to collect actual production data; The variables of the collected actual production data are screened according to the correlation between the variables and the convexity, and a preset number of mechanical property variable data with the highest correlation with the convexity of the strip of the finishing mill in the hot rolling production process are selected.

3. The method for real-time prediction of strip convexity of a finishing mill group in a hot rolling production process according to claim 1, characterized in that: The sliding window operation is based on which the acquired mechanical variable data is reconstructed through a sliding window, comprising: The mechanical variable data is represented as P*N dimensional data with the value of each mechanical variable as the vertical axis and the data time series as the horizontal axis; where N represents the number of mechanical variables and P represents the length of the time series; A sliding window operation is performed on the vertical axis data at each time step to reconstruct the data at each time step into multiple subsequences composed of M features; where M is the sliding window length.

4. The method for real-time prediction of strip convexity of a finishing mill group in a hot rolling production process according to claim 1, characterized in that: The improvement of the convolution gated recurrent unit includes: Combine the convolutional block attention module with the convolutional layer in the convolutional gated recurrent unit.

5. The method for real-time prediction of strip convexity of a finishing mill group in a hot rolling production process according to claim 4, characterized in that: When the improved convolutional gated recurrent unit performs feature extraction, the fully connected layer in the convolutional gated recurrent unit is first replaced by the convolutional layer to enhance the local feature extraction capability, and then the convolutional block attention module enhances the expression capability of the convolutional layer to further improve the local feature extraction capability of the model.

6. The method for real-time prediction of strip convexity of a finishing mill group in a hot rolling production process according to claim 1, characterized in that: After realizing the strip convexity prediction by using the trained strip convexity prediction model of the finishing mill in the hot rolling production process, the real-time prediction method for the strip convexity of the finishing mill in the hot rolling production process further comprises: The model prediction ability is evaluated using preset evaluation indicators.

7. The method for real-time prediction of strip convexity of a finishing mill group in a hot rolling production process according to claim 6, characterized in that: The preset evaluation index is a combination of one or more of the root mean square error RMSE, the mean absolute error MAE, the mean relative error MRE and the determination coefficient R2.

8. A real-time prediction device for the convexity of strip steel in a finishing mill during hot rolling production, characterized in that: include: Data processing module for: Obtain data on multiple mechanical variables that affect the strip crown of the finishing mill in the hot rolling production process; Based on the sliding window operation, the acquired mechanical variable data is reconstructed by sliding window; Construct a sample data set using the mechanical variable data reconstructed by the sliding window; A model building module is used to improve the convolution gated cyclic unit and use the improved convolution gated cyclic unit to build a strip crown prediction model for the finishing mill in the hot rolling production process; A model training module, used to train the strip convexity prediction model of the finishing mill group in the hot rolling production process constructed by the model construction module using the sample data set constructed by the data processing module; The module for real-time prediction of strip convexity of a finishing mill in a hot rolling production process is used to realize strip convexity prediction by using the strip convexity prediction model of a finishing mill in a hot rolling production process trained by the model training module.

Images