A method and apparatus for predicting rolling force in cold continuous rolling mills based on IDBO-SA-LSTM

Abstract

This application relates to the field of rolling force prediction technology, and particularly to a method and apparatus for predicting rolling force in cold continuous rolling mills based on IDBO-SA-LSTM. The method includes: optimizing a dung beetle optimization algorithm based on a preset rolling force calculation model, a Circle chaotic mapping strategy, a golden sine strategy, and dynamic weight coefficients to construct an improved dung beetle optimization algorithm; performing performance tests on the improved dung beetle optimization algorithm using various benchmark test functions to obtain an improved dung beetle optimization algorithm that meets preset performance test requirements; automatically optimizing a preset adaptive long short-term memory network based on the improved dung beetle optimization algorithm, and constructing a target SA-LSTM rolling force prediction model using the automatically optimized adaptive long short-term memory network, so as to use the target SA-LSTM rolling force prediction model to predict the rolling force in real time during the online prediction stage. This solves the problem that existing mechanistic models contain many difficult-to-determine empirical parameters, which cannot guarantee the generalization performance and accuracy of cold continuous rolling mill rolling force prediction.

Description

Technical Field

[0001] This application relates to the field of cold continuous rolling force prediction technology, and in particular to a method and apparatus for predicting cold continuous rolling force based on IDBO-SA-LSTM. Background Technology

[0002] In recent years, the new energy industry has developed rapidly, and the demand and output of aluminum foil, which is widely used in the production and manufacturing of lithium batteries, are also constantly increasing. Cold rolling is one of the important processes in the production of aluminum foil. During the cold rolling process, rolling force is the key means to control the thickness and shape of aluminum foil products. Therefore, the rolling force calculation model has an important impact on the quality of aluminum foil products.

[0003] However, in actual production, rolling force is affected by a variety of on-site factors. Traditional mechanism models often contain many difficult-to-determine empirical parameters, making it difficult to guarantee the generalization performance and accuracy of the model, which urgently needs to be addressed. Summary of the Invention

[0004] This application provides a method and apparatus for predicting cold rolling force based on IDBO-SA-LSTM, in order to solve the problems that existing mechanism models contain many difficult-to-determine empirical parameters, which cannot guarantee the generalization performance and accuracy of cold rolling force prediction.

[0005] The first aspect of this application provides a method for predicting rolling force in cold continuous rolling mills based on IDBO-SA-LSTM, applied in the offline training stage, including the following steps: optimizing a preset dung beetle optimization algorithm based on a preset rolling force calculation model, a Circle chaotic mapping strategy, a golden sine strategy, and dynamic weight coefficients to construct an improved dung beetle optimization algorithm; performing performance tests on the improved dung beetle optimization algorithm through a variety of preset benchmark test functions to obtain an improved dung beetle optimization algorithm that meets preset performance test requirements; automatically optimizing a preset adaptive long short-term memory network based on the improved dung beetle optimization algorithm, and constructing a target SA-LSTM rolling force prediction model through the automatically optimized adaptive long short-term memory network, so as to use the target SA-LSTM rolling force prediction model to predict the rolling force in real time during the online prediction stage.

[0006] Optionally, in one embodiment of this application, the step of performing performance testing on the improved dung beetle optimization algorithm using a plurality of preset benchmark testing functions to obtain an improved dung beetle optimization algorithm that meets preset performance testing requirements includes: determining the plurality of benchmark testing functions, and setting the population size and maximum number of iterations for each of the plurality of benchmark testing functions in a single iteration, wherein the plurality of benchmark testing functions includes a plurality of single-peak testing functions and a plurality of multi-peak testing functions; and performing performance testing on the improved dung beetle optimization algorithm using each of the benchmark testing functions based on the population size in a single iteration and the maximum number of iterations.

[0007] Optionally, in one embodiment of this application, the step of automatically optimizing a preset adaptive long short-term memory network based on the improved dung beetle optimization algorithm, and constructing a target SA-LSTM rolling force prediction model through the automatically optimized adaptive long short-term memory network, includes: constructing the adaptive long short-term memory network based on a preset self-attention mechanism, LSTM layer, activation layer, fully connected layer, and regression layer; acquiring raw cold rolling force data, and randomly shuffling and normalizing the raw cold rolling force data to construct a target training format dataset, and dividing the target training format dataset to generate a training dataset and a test dataset; initializing multiple parameters in the improved dung beetle optimization algorithm, and initializing the dung beetle corresponding to the improved dung beetle optimization algorithm based on the Circle chaotic mapping strategy. A mantis population is used to obtain multiple individual locations, and the adaptive long short-term memory network is trained using these multiple individual locations to generate individual locations with target fitness. The multiple parameters include population size, maximum number of iterations, number of hidden layer neurons, initial learning rate, regularization coefficient, and the proportions of behaviors such as ball rolling, reproduction, foraging, and theft within the population. The improved mantis optimization algorithm is used to update the individual locations with target fitness, iterating up to the maximum number of iterations to obtain optimal solution parameters. The adaptive long short-term memory network is then trained using the optimal solution parameters and the training dataset to obtain an initial SA-LSTM rolling force prediction model. The initial SA-LSTM rolling force prediction model is then fine-tuned using the test dataset to obtain the target SA-LSTM rolling force prediction model.

[0008] A second aspect of this application provides a method for predicting rolling force in cold continuous rolling mills based on IDBO-SA-LSTM, applied in the online prediction stage. The method includes: acquiring actual data during the smooth rolling process of a target cold continuous rolling mill; normalizing and calculating the fitness of the actual data to obtain a corresponding fitness calculation index; inputting the actual data into a pre-constructed target SA-LSTM rolling force prediction model, and combining the fitness calculation index to output the rolling force prediction result during the smooth rolling process of the target cold continuous rolling mill. The target SA-LSTM rolling force prediction model is obtained by training an adaptive long short-term memory network optimized using a pre-defined improved dung beetle optimization algorithm on a preset training dataset and test dataset of cold continuous rolling force.

[0009] A third aspect of this application provides a cold continuous rolling force prediction device based on IDBO-SA-LSTM, applied in the offline training stage, comprising the following steps: an optimization module, used to optimize a preset dung beetle optimization algorithm based on a preset rolling force calculation model, a Circle chaotic mapping strategy, a golden sine strategy, and dynamic weight coefficients, to construct an improved dung beetle optimization algorithm; a testing module, used to perform performance testing on the improved dung beetle optimization algorithm through a variety of preset benchmark testing functions, to obtain an improved dung beetle optimization algorithm that meets preset performance testing requirements; and a modeling module, used to automatically optimize a preset adaptive long short-term memory network based on the improved dung beetle optimization algorithm, and construct a target SA-LSTM rolling force prediction model through the automatically optimized adaptive long short-term memory network, so as to use the target SA-LSTM rolling force prediction model to predict the rolling force in real time during the online prediction stage.

[0010] Optionally, in one embodiment of this application, the testing module includes: a setting unit, configured to determine the plurality of benchmark testing functions and set the population size and maximum number of iterations per single iteration for each of the plurality of benchmark testing functions, wherein the plurality of benchmark testing functions includes a plurality of unimodal testing functions and a plurality of multimodal testing functions; and an analysis unit, configured to perform performance testing on the improved dung beetle optimization algorithm using each of the benchmark testing functions based on the population size per single iteration and the maximum number of iterations.

[0011] Optionally, in one embodiment of this application, the modeling module includes: a construction unit, used to construct the adaptive long short-term memory network based on a preset self-attention mechanism, LSTM layer, activation layer, fully connected layer, and regression layer; a partitioning unit, used to acquire raw data of cold continuous rolling force, and to perform random shuffling and normalization operations on the raw data of cold continuous rolling force to construct a target training format dataset, and to partition the target training format dataset to generate a training dataset and a test dataset; and an initialization unit, used to initialize multiple parameters in the improved dung beetle optimization algorithm, and to initialize the dung beetle population corresponding to the improved dung beetle optimization algorithm based on the Circle chaotic mapping strategy to obtain multiple individual positions, and to train the self-attention network using the multiple individual positions. An adaptive long short-term memory network is used to generate individual positions for the target fitness, wherein the multiple parameters include population size, maximum number of iterations, number of hidden layer neurons, initial learning rate, regularization coefficient, and the proportion of behaviors such as ball rolling, reproduction, foraging, and stealing in the population; an iterative unit is used to update the individual positions for the target fitness using the improved dung beetle optimization algorithm and iterate cyclically to the maximum number of iterations to obtain the optimal solution parameters, and train the adaptive long short-term memory network using the optimal solution parameters and the training dataset to obtain an initial SA-LSTM rolling force prediction model; a fine-tuning unit is used to fine-tune the initial SA-LSTM rolling force prediction model using the test dataset to obtain the target SA-LSTM rolling force prediction model.

[0012] The fourth aspect of this application provides a cold rolling mill rolling force prediction device based on IDBO-SA-LSTM, applied in the online prediction stage, comprising: an acquisition module for acquiring actual data during the smooth rolling process of a target cold rolling mill; a calculation module for normalizing and calculating the fitness of the actual data to obtain a corresponding fitness calculation index; and a prediction module for inputting the actual data into a pre-constructed target SA-LSTM rolling force prediction model and combining it with the fitness calculation index to output the rolling force prediction result during the smooth rolling process of the target cold rolling mill, wherein the target SA-LSTM rolling force prediction model is obtained by training an adaptive long short-term memory network optimized by a pre-defined improved dung beetle optimization algorithm using a preset training dataset and a test dataset of cold rolling forces.

[0013] A fifth aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the cold continuous rolling force prediction method based on IDBO-SA-LSTM as described in the above embodiments.

[0014] A sixth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predicting cold continuous rolling force based on IDBO-SA-LSTM.

[0015] A seventh aspect of this application provides a computer program product, including a computer program that is executed to implement the above-described method for predicting cold continuous rolling force based on IDBO-SA-LSTM.

[0016] Therefore, the embodiments of this application have the following beneficial effects:

[0017] The embodiments of this application optimize a pre-defined dung beetle optimization algorithm based on a pre-defined rolling force calculation model, a Circle chaotic mapping strategy, a golden sine strategy, and dynamic weight coefficients to construct an improved dung beetle optimization algorithm. The improved dung beetle optimization algorithm is then tested using various pre-defined benchmark functions to obtain an improved algorithm that meets the pre-defined performance test requirements. Based on the improved dung beetle optimization algorithm, a pre-defined adaptive long short-term memory network is automatically optimized, and a target SA-LSTM rolling force prediction model is constructed using the automatically optimized adaptive long short-term memory network. This model is then used for real-time prediction of rolling force during the online prediction phase. This application optimizes the long short-term memory network combined with a self-attention mechanism using an improved dung beetle optimization algorithm, effectively improving the accuracy and generalization performance of cold continuous rolling force prediction, and achieving high-precision rolling force prediction in the aluminum foil cold continuous rolling process. This solves the problem that existing mechanistic models contain many difficult-to-determine empirical parameters, which cannot guarantee the generalization performance and accuracy of cold continuous rolling force prediction.

[0018] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0019] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0020] Figure 1 This is a flowchart of a cold continuous rolling force prediction method based on IDBO-SA-LSTM applied in the offline training stage, according to an embodiment of this application.

[0021] Figure 2 A schematic diagram of an aluminum foil cold continuous rolling production equipment provided for one embodiment of this application;

[0022] Figure 3A schematic diagram illustrating the distribution of a Circle chaotic mapping iteration after 500 iterations, provided as an embodiment of this application;

[0023] Figure 4 (a) in the figure is a schematic diagram of the optimization process curve of the F1(x) test function provided in an embodiment of this application;

[0024] Figure 4 (b) is a schematic diagram of the optimization process curve of the F2(x) test function provided in an embodiment of this application;

[0025] Figure 4 (c) is a schematic diagram of the optimization process curve of the F3(x) test function provided in an embodiment of this application;

[0026] Figure 4 (d) is a schematic diagram of the optimization process curve of the F4(x) test function provided in an embodiment of this application;

[0027] Figure 4 (e) in the figure is a schematic diagram of the optimization process curve of the F5(x) test function provided in an embodiment of this application;

[0028] Figure 4 (f) is a schematic diagram of the optimization process curve of the F6(x) test function provided in an embodiment of this application;

[0029] Figure 5 A schematic diagram of the execution logic for cold continuous rolling force prediction based on IDBO-SA-LSTM is provided for one embodiment of this application;

[0030] Figure 6 A schematic diagram of a deep learning network structure for a rolling force prediction model provided in one embodiment of this application;

[0031] Figure 7 This is a flowchart of a cold continuous rolling force prediction method based on IDBO-SA-LSTM applied in the online prediction stage, according to an embodiment of this application.

[0032] Figure 8 A comparison chart of predicted and actual rolling force values ​​for a single pass is provided as an embodiment of this application;

[0033] Figure 9 A schematic diagram showing the comparison between predicted and actual rolling force values ​​for two passes, provided as an embodiment of this application;

[0034] Figure 10 This is an example diagram of a cold continuous rolling force prediction device based on IDBO-SA-LSTM applied in the offline training phase according to an embodiment of this application.

[0035] Figure 11 This is an example diagram of a cold continuous rolling force prediction device based on IDBO-SA-LSTM applied in the online prediction stage according to an embodiment of this application;

[0036] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0037] Among them, 10-a cold rolling force prediction device based on IDBO-SA-LSTM applied to the offline training stage, 20-a cold rolling force prediction device based on IDBO-SA-LSTM applied to the online prediction stage; 101-optimization module, 102-testing module, 103-modeling module; 201-acquisition module, 202-calculation module, 203-prediction module; 1201-memory, 1202-processor, 1203-communication interface. Detailed Implementation

[0038] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0039] The following description, with reference to the accompanying drawings, describes a method and apparatus for predicting rolling force in cold continuous rolling mills based on IDBO-SA-LSTM, according to embodiments of this application. Addressing the problems mentioned in the background section, this application provides a method for predicting rolling force in cold continuous rolling mills based on IDBO-SA-LSTM. In this method, a preset dung beetle optimization algorithm is optimized using a preset rolling force calculation model, a Circle chaotic mapping strategy, a golden sine strategy, and dynamic weight coefficients to construct an improved dung beetle optimization algorithm. The improved dung beetle optimization algorithm is then tested using multiple preset benchmark functions to obtain an improved dung beetle optimization algorithm that meets preset performance test requirements. Based on the improved dung beetle optimization algorithm, a preset adaptive long short-term memory network is automatically optimized, and a target SA-LSTM rolling force prediction model is constructed using the automatically optimized adaptive long short-term memory network. This target SA-LSTM rolling force prediction model is then used to predict the rolling force in real time during the online prediction phase. This application improves the dung beetle optimization algorithm to optimize the long short-term memory network combined with the self-attention mechanism, effectively enhancing the accuracy and generalization performance of cold continuous rolling force prediction, and achieving high-precision rolling force prediction for the aluminum foil cold continuous rolling process. This solves the problem that existing mechanistic models contain many difficult-to-determine empirical parameters, which cannot guarantee the generalization performance and accuracy of cold continuous rolling force prediction.

[0040] Specifically, Figure 1 This is a flowchart illustrating a cold continuous rolling force prediction method based on IDBO-SA-LSTM applied in the offline training phase, as provided in an embodiment of this application.

[0041] like Figure 1 As shown, the IDBO-SA-LSTM-based cold continuous rolling force prediction method applied in the offline training phase includes the following steps:

[0042] In step S101, the preset dung beetle optimization algorithm is optimized based on the preset rolling force calculation model, Circle chaotic mapping strategy, golden sine strategy and dynamic weight coefficients to construct an improved dung beetle optimization algorithm.

[0043] In step S102, the performance of the improved dung beetle optimization algorithm is tested using a variety of preset benchmark test functions to obtain an improved dung beetle optimization algorithm that meets the preset performance test requirements.

[0044] The embodiments of this application first construct a rolling force calculation model. Those skilled in the art should understand that commonly used cold rolling force calculation models include the Zelikoff formula, the Stone formula, the Hill formula, etc., among which the most commonly used is the Hill formula simplified by Brand and Ford, as shown in equation (1):

[0045]

[0046] Where P is the average rolling force on the workpiece, B is the width of the workpiece, and l c Q is the contact arc length between the workpiece and the roll. P K is the stress state coefficient related to the friction coefficient. T The tension influence coefficient related to the tension of the rolled piece. The average deformation resistance is related to the chemical properties of the rolled piece.

[0047] In the embodiments of this application, the layout of the aluminum foil cold rolling production line is as follows: Figure 2 As shown, in actual production, rolling conditions change with the rolling process. The rolling force in each pass is related not only to process parameters such as workpiece tension, rolling speed, and reduction, but also to boundary conditions such as roll wear, cooling, and lubrication. In the theoretical calculation model of rolling force, the friction coefficient and average deformation resistance are difficult to measure, resulting in low calculation accuracy and generalization performance of the theoretical model in actual production.

[0048] Furthermore, when using traditional DBO to optimize the hyperparameters of the SA-LSTM model, it was found that the algorithm is prone to getting trapped in local optima during computation. To improve computational accuracy, this application proposes an improved Dung Beetle Optimizer (IDBO) algorithm. This algorithm increases the uniformity of the initial population distribution through Circle chaotic mapping, enhances the global search performance of DBO using a golden sine strategy, and introduces dynamic weight coefficients to balance the global and local search operations. The specific details are as follows:

[0049] 1. Circle chaotic mapping:

[0050] The original Dung Beetle Optimizer (DBO) algorithm initializes the population randomly, which cannot guarantee an even distribution of the initial population in the search space, resulting in low diversity of the initial population. Chaotic mapping, with its non-repetitive, random, and chaotic ergodic properties, can achieve a more uniform population distribution compared to probability-dependent random generation. Therefore, embodiments of this application can utilize chaotic mapping to generate the initial population to increase the diversity of potential solutions.

[0051] To enhance the stability of the DBO algorithm and the uniformity of the initial population distribution, this patent uses Circle chaotic mapping instead of random numbers to initialize the dung beetle population. The specific calculation method of Circle chaotic mapping is shown in equation (2):

[0052]

[0053] Where, x n This represents the magnitude of the result in the nth iteration, where a and b are both constants.

[0054] As one possible approach, the distribution of the Circle chaotic mapping after 500 iterations in this embodiment is as follows: Figure 3 As shown, by Figure 3 It can be seen that the distribution of the Circle chaotic map in space is relatively stable and uniform, and it has good ergodicity.

[0055] 2. Golden Sine Strategy

[0056] Golden-SA is a novel metaheuristic algorithm that uses a sine function for iterative optimization, which can accelerate convergence and enhance global search capabilities. Its update process is shown in equation (3):

[0057]

[0058] in, and Let R1 and R2 represent the spatial position and optimal position of the i-th individual in the t-th iteration, respectively. Let R1 be a random number representing the distance the individual moves, and R2 be a random number representing the direction the individual moves. Let x1 and x2 be two constant coefficients calculated from the golden ratio.

[0059] In the dancing behavior of the dung beetle, the embodiments of this application can determine the direction of movement in the next iteration based on the environment surrounding the current position. However, if the current position is already a local optimum, the dung beetle's dancing movements will change very little, easily causing the population to get trapped in a local optimum. Therefore, to enhance the global search capability of the DBO algorithm, the embodiments of this application can use Golden-SA to expand the search space and replace the dancing update strategy when the dung beetle has no obstacles. At the same time, R1 and R2 can effectively control the search direction and search step size of the dung beetle population, further balancing the algorithm's global search capability and local search capability.

[0060] 3. Dynamic weighting coefficients

[0061] In the dung beetle theft behavior, the position update strategy is significantly influenced by the global optimal position. Since the population often approaches the optimal solution range in the later stages of iteration, focusing on local searches can improve the accuracy of the search results. To further improve the solution accuracy and reduce unnecessary divergence, this application can introduce dynamic weight coefficients to optimize the dung beetle theft behavior and enhance the local search capability of the algorithm in the later stages of iteration. The optimized dung beetle theft update strategy is shown in equations (4)-(6):

[0062] x i t+1 =k1×X b +k2×S×g×(|x i t -X * |+|x i t -X b |) (4)

[0063]

[0064] Here, k1 and k2 are dynamic factors determined by the number of iterations, which can enhance the local exploitation capability of the algorithm in the later stages of iteration.

[0065] Therefore, the embodiments of this application can optimize the dung beetle optimization algorithm through the Circle chaotic mapping strategy, the golden sine strategy, and dynamic weight coefficients to construct an improved dung beetle optimization algorithm; subsequently, the embodiments of this application can perform performance testing on the improved dung beetle optimization algorithm.

[0066] Optionally, in one embodiment of this application, the improved dung beetle optimization algorithm is tested for performance using a variety of preset benchmark testing functions to obtain an improved dung beetle optimization algorithm that meets preset performance testing requirements. This includes: determining a variety of benchmark testing functions, and setting the population size and maximum number of iterations for each test function in a single iteration, wherein the variety of benchmark testing functions includes a variety of single-peak test functions and a variety of multi-peak test functions; and testing the performance of the improved dung beetle optimization algorithm using each test function based on the population size and maximum number of iterations in a single iteration.

[0067] To test the optimization performance of the IDBO algorithm, this application uses a benchmark test function to test its performance and compares it with the Sparrow Search Algorithm (SSA), Grey Wolf Optimizer (GWO), and DBO algorithm. Basic information about the benchmark test function is shown in Table 1.

[0068] Table 1

[0069]

[0070] It should be noted that the test functions F1 to F4 in Table 1 are unimodal functions, while F5 and F6 are multimodal functions. When running the test functions, the population size for each iteration is set to 30, and the maximum number of iterations is set to 1000. The optimization process of each search algorithm on the baseline function is as follows: Figure 4 (a)- Figure 4 As shown in (f) in the figure, by Figure 4 (a)- Figure 4 As shown in (f), the IDBO algorithm has a faster convergence speed and higher optimization accuracy than the other three algorithms.

[0071] In step S103, based on the improved dung beetle optimization algorithm, the preset adaptive long short-term memory network is automatically optimized, and the target SA-LSTM rolling force prediction model is constructed through the automatically optimized adaptive long short-term memory network, so as to use the target SA-LSTM rolling force prediction model to predict the rolling force in real time during the online prediction stage.

[0072] Furthermore, embodiments of this application also require optimization of the Self-Attention-Long Short-Term Memory (SA-LSTM) network based on the improved dung beetle optimization algorithm to construct a target SA-LSTM rolling force prediction model suitable for high-precision rolling force prediction in the cold continuous rolling process of aluminum foil.

[0073] Optionally, in one embodiment of this application, an improved dung beetle optimization algorithm is used to automatically optimize a preset adaptive long short-term memory network, and a target SA-LSTM rolling force prediction model is constructed using the automatically optimized adaptive long short-term memory network. This includes: constructing an adaptive long short-term memory network based on a preset self-attention mechanism, LSTM layers, activation layers, fully connected layers, and regression layers; acquiring raw cold rolling force data, and randomly shuffling and normalizing the raw cold rolling force data to construct a target training format dataset, and dividing the target training format dataset to generate a training dataset and a test dataset; initializing multiple parameters in the improved dung beetle optimization algorithm, and initializing the improved dung beetle optimization algorithm based on the Circle chaotic mapping strategy. The method involves using a dung beetle population to obtain multiple individual locations, and then training an adaptive long short-term memory network (LSTM) to generate the target fitness individual locations. Several parameters are used, including population size, maximum number of iterations, number of hidden layer neurons, initial learning rate, regularization coefficient, and the proportions of rolling, breeding, foraging, and stealing behaviors within the population. The target fitness individual locations are updated using an improved dung beetle optimization algorithm, iterated up to the maximum number of iterations to obtain the optimal solution parameters. The LSTM network is then trained using these optimal parameters and the training dataset to obtain an initial SA-LSTM rolling force prediction model. Finally, the initial SA-LSTM rolling force prediction model is fine-tuned using a test dataset to obtain the target SA-LSTM rolling force prediction model.

[0074] To fully extract the deep coupling relationships between variables during cold rolling, embodiments of this application can construct, as follows: Figure 5 The deep learning model shown predicts rolling force. To avoid overfitting and improve the model's running speed, this embodiment first uses a self-attention mechanism to extract features from the original input, then uses an LSTM layer to further fit the nonlinear relationship between variables, and then maps the output of the LSTM layer to the dimension space where the rolling force is located through activation layers, fully connected layers, and regression layers, thereby fully fitting the coupling between rolling force and industrial field parameters.

[0075] Furthermore, various hyperparameters included in the LSTM layer significantly impact the model's prediction results. The number of hidden layer neurons (N), the initial learning rate (a), and the regularization coefficient (L) are all crucial hyperparameters. The number of hidden layer neurons (N) determines the capacity of the LSTM model. If N is too small, the network may not be able to store enough state information, thus failing to effectively handle nonlinear coupling relationships in the input sequence. If N is too large, it can lead to overfitting, reducing the model's accuracy and generalization performance. The initial learning rate (a) refers to the initial step size used by the optimization algorithm to update the model parameters at the start of training. If a is too small, the model training speed will be too slow, potentially causing the optimizer to stagnate near local optima. If a is too large, the model may fail to converge. The regularization coefficient (L) is a penalty term added to the loss function to prevent the model from becoming too complex or overfitting the training data. If L is too small, the impact of the regularization term on the loss function will be weakened, easily leading to overfitting. If L is too large, it will excessively restrict the model's parameter fitting ability, resulting in decreased prediction performance.

[0076] However, manually adjusting parameters requires significant labor costs. Therefore, this embodiment of the application uses the IDBO algorithm to automatically optimize these hyperparameters of the LSTM model, constructing a model such as... Figure 6 The IDBO-SA-LSTM model shown predicts rolling force. The specific prediction process of this IDBO-SA-LSTM model is as follows:

[0077] (1) Read the raw data of cold continuous rolling force, randomly shuffle and normalize it to construct a trainable dataset, and divide it into training set and test set according to the ratio of 8:2;

[0078] (2) Initialize the IDBO algorithm parameters, setting the population size to 20, the maximum number of iterations to 50, the ratio of rolling, breeding, foraging, and stealing behaviors in the population to 2:1:1:1, the optimization range of the number of hidden layer neurons N to [10, 100], and the optimization range of the initial learning rate a to [1×10]. -5 1×10 -1 The optimization range of the regularization coefficient L is [1×10]. -8 1×10 -1 ];

[0079] (3) The dung beetle population is initialized using the Circle chaotic mapping. Each individual represents a set of feasible solutions. All individual positions are substituted into the SA-LSTM model for training to obtain the individual position with the best fitness.

[0080] (4) The DBO algorithm is optimized using the golden sine strategy and dynamic weight coefficients to obtain the IDBO algorithm, and the individual position is updated according to the IDBO algorithm.

[0081] (5) Iterate until the maximum number of iterations, and then substitute the obtained optimal solution parameters into the SA-LSTM model for training to obtain the optimal SA-LSTM rolling force prediction model (i.e. the target SA-LSTM rolling force prediction model).

[0082] Furthermore, in the subsequent online prediction stage, the target IDBO-SA-LSTM model can be used to predict the rolling force, and the prediction results can be analyzed to improve the calculation accuracy and generalization performance of the model.

[0083] The cold continuous rolling force prediction method based on IDBO-SA-LSTM proposed in this application, applied to the offline training stage, optimizes a preset dung beetle optimization algorithm based on a preset rolling force calculation model, Circle chaotic mapping strategy, golden sine strategy, and dynamic weight coefficients to construct an improved dung beetle optimization algorithm. The improved dung beetle optimization algorithm is then tested using multiple preset benchmark functions to obtain an improved algorithm that meets preset performance test requirements. Based on the improved dung beetle optimization algorithm, a preset adaptive long short-term memory network is automatically optimized, and a target SA-LSTM rolling force prediction model is constructed using the automatically optimized adaptive long short-term memory network. This model is then used for real-time prediction of rolling force during the online prediction stage. This application optimizes the long short-term memory network combined with a self-attention mechanism using an improved dung beetle optimization algorithm, effectively improving the accuracy and generalization performance of cold continuous rolling force prediction, and achieving high-precision rolling force prediction in the aluminum foil cold continuous rolling process.

[0084] Figure 7 A flowchart illustrating a cold continuous rolling force prediction method based on IDBO-SA-LSTM applied in the online prediction stage, provided as an embodiment of this application.

[0085] like Figure 7 As shown, the cold continuous rolling force prediction method based on IDBO-SA-LSTM includes the following steps:

[0086] In step S701, actual data during the smooth rolling process of the target cold rolling mill are obtained.

[0087] In step S702, the actual data is normalized and fitness is calculated to obtain the corresponding fitness calculation index.

[0088] In step S703, the actual data is input into the pre-constructed target SA-LSTM rolling force prediction model, and combined with the fitness calculation index, the rolling force prediction result of the target cold continuous rolling mill during the smooth rolling process is output. The target SA-LSTM rolling force prediction model is obtained by training an adaptive long short-term memory network optimized by a pre-defined improved dung beetle optimization algorithm on the pre-defined training dataset and test dataset of the cold continuous rolling force.

[0089] The following detailed explanation of the process of predicting cold rolling force based on IDBO-SA-LSTM is provided through a specific embodiment.

[0090] 1. Hardware platform setup:

[0091] Regarding the operating system and computer hardware, specific embodiments of this application can perform model training under a Windows 10 Professional 64-bit operating system, with an Intel(R) Xeon(R) Gold 5218 CPU @ 2.30GHz processor, 256GB of memory, and an NVIDIA TITAN RTX graphics card. Regarding software, embodiments of this application use MATLAB R2024b to build the network structure and conduct experiments.

[0092] 2. Data selection:

[0093] A specific embodiment of this application can monitor a 3000 series aluminum foil production line in a factory, using actual data from the stable rolling process of a 1950 mm two-stand six-roll cold continuous rolling mill as the data source. Data is collected from 109 stable rolling processes under different production conditions, resulting in 6554 sets of raw data. Some of the raw data are shown in Table 2.

[0094] Table 2

[0095]

[0096] As shown in Table 2, the embodiments of this application can use variables such as exit speed, air system pressure, work roll bending force, support roll bending force, shift position, inlet thickness difference, inlet tension, intermediate tension, exit tension, absolute position, and thickness error as input variables, and rolling force as output variables.

[0097] 3. Normalization method and fitness calculation method:

[0098] In a specific embodiment of this application, the original data can be mapped to the [0,1] interval using the maximum-minimum value normalization method. The specific calculation method is shown in equation (7):

[0099]

[0100] Where x represents the original data, x max x represents the maximum value in the original data. min x is the minimum value in the original data. norm This represents the value of the original data after normalization.

[0101] In the process of using IDBO for optimization, the root mean squared error (RMSE) can be used as the fitness calculation index in this embodiment of the application. The specific calculation method is shown in Equation (8):

[0102]

[0103] in, x represents the predicted rolling force. i This indicates the actual value of the rolling force.

[0104] Through experiments, the embodiments of this application show the following prediction results of the IDBO-SA-LSTM model for two rolling forces: Figure 8 and Figure 9 As shown, by Figure 8 and Figure 9 It can be seen that the probability that the model's rolling force prediction error is within the 4% error band in both passes is greater than 99%.

[0105] The cold continuous rolling force prediction method based on IDBO-SA-LSTM proposed in this application, applied to the online prediction stage, obtains actual data during the smooth rolling process of the target cold continuous rolling mill; normalizes and calculates the fitness of the actual data to obtain the corresponding fitness calculation index; inputs the actual data into a pre-constructed target SA-LSTM rolling force prediction model, and combines it with the fitness calculation index to output the rolling force prediction result during the smooth rolling process of the target cold continuous rolling mill. The target SA-LSTM rolling force prediction model is obtained by training an adaptive long short-term memory network optimized by a pre-defined improved dung beetle optimization algorithm using a pre-defined training dataset and test dataset of cold continuous rolling force. This application optimizes the long short-term memory network combined with a self-attention mechanism using an improved dung beetle optimization algorithm, effectively improving the accuracy and generalization performance of cold continuous rolling force prediction, and achieving high-precision rolling force prediction for the aluminum foil cold continuous rolling process.

[0106] Secondly, with reference to the accompanying drawings, a cold continuous rolling force prediction device based on IDBO-SA-LSTM according to an embodiment of this application is described.

[0107] Figure 10 This is a block diagram of a cold continuous rolling force prediction device based on IDBO-SA-LSTM applied to the offline training phase according to an embodiment of this application.

[0108] like Figure 10 As shown, the IDBO-SA-LSTM-based cold rolling force prediction device 10 applied in the offline training phase includes: an optimization module 101, a testing module 102, and a modeling module 103.

[0109] The optimization module 101 is used to optimize the preset dung beetle optimization algorithm based on the preset rolling force calculation model, Circle chaotic mapping strategy, golden sine strategy and dynamic weight coefficients, so as to construct an improved dung beetle optimization algorithm.

[0110] The testing module 102 is used to perform performance tests on the improved dung beetle optimization algorithm through a variety of preset benchmark testing functions, so as to obtain an improved dung beetle optimization algorithm that meets the preset performance test requirements.

[0111] Modeling module 103 is used to automatically optimize a preset adaptive long short-term memory network based on an improved dung beetle optimization algorithm, and to construct a target SA-LSTM rolling force prediction model through the automatically optimized adaptive long short-term memory network, so as to use the target SA-LSTM rolling force prediction model to predict the rolling force in real time during the online prediction stage.

[0112] Optionally, in one embodiment of this application, the test module 200 includes a setting unit and an analysis unit.

[0113] The setting unit is used to determine multiple benchmark testing functions and set the population size and maximum number of iterations for each benchmark testing function in a single iteration. The multiple benchmark testing functions include multiple unimodal testing functions and multiple multimodal testing functions.

[0114] The analysis unit is used to perform performance testing on the improved dung beetle optimization algorithm using each test function, based on the population size and maximum number of iterations in a single iteration.

[0115] Optionally, in one embodiment of this application, the modeling module 300 includes: a construction unit, a partitioning unit, an initialization unit, an iteration unit, and a fine-tuning unit.

[0116] The building unit is used to construct an adaptive long short-term memory network based on a preset self-attention mechanism, LSTM layer, activation layer, fully connected layer and regression layer.

[0117] The unit is used to obtain the raw data of cold rolling force, and to randomly shuffle and normalize the raw data of cold rolling force to construct the target training format dataset. The target training format dataset is then divided to generate the training dataset and the test dataset.

[0118] An initialization unit is used to initialize multiple parameters in the improved dung beetle optimization algorithm and, based on the Circle chaotic mapping strategy, initialize the dung beetle population corresponding to the improved dung beetle optimization algorithm to obtain multiple individual positions. The adaptive long short-term memory network is then trained using these multiple individual positions to generate the individual positions of the target fitness. The multiple parameters include the population size, maximum number of iterations, number of hidden layer neurons, initial learning rate, regularization coefficient, and the proportion of behaviors such as rolling, breeding, foraging, and stealing in the population.

[0119] The iterative unit is used to update the individual position of the target fitness using the improved dung beetle optimization algorithm and iterates until the maximum number of iterations to obtain the optimal solution parameters. The adaptive long short-term memory network is then trained using the optimal solution parameters and the training dataset to obtain the initial SA-LSTM rolling force prediction model.

[0120] The fine-tuning unit is used to fine-tune the initial SA-LSTM rolling force prediction model using the test dataset to obtain the target SA-LSTM rolling force prediction model.

[0121] It should be noted that the foregoing explanation of the embodiment of the cold continuous rolling force prediction method based on IDBO-SA-LSTM applied to the offline training stage also applies to the cold continuous rolling force prediction device based on IDBO-SA-LSTM applied to the offline training stage of this embodiment, and will not be repeated here.

[0122] The cold continuous rolling force prediction device based on IDBO-SA-LSTM proposed in this application includes an optimization module 101, which optimizes a preset dung beetle optimization algorithm based on a preset rolling force calculation model, a Circle chaotic mapping strategy, a golden sine strategy, and dynamic weight coefficients to construct an improved dung beetle optimization algorithm; a testing module 102, which performs performance tests on the improved dung beetle optimization algorithm through a variety of preset benchmark test functions to obtain an improved dung beetle optimization algorithm that meets preset performance test requirements; and a modeling module 103, which automatically optimizes a preset adaptive long short-term memory network based on the improved dung beetle optimization algorithm, and constructs a target SA-LSTM rolling force prediction model through the automatically optimized adaptive long short-term memory network, so as to use the target SA-LSTM rolling force prediction model to predict the rolling force in real time during the online prediction stage. This application improves the dung beetle optimization algorithm to optimize the long short-term memory network combined with the self-attention mechanism, which effectively improves the accuracy and generalization performance of cold continuous rolling force prediction and realizes high-precision rolling force prediction in the cold continuous rolling process of aluminum foil.

[0123] Figure 11 This is a block diagram of a cold continuous rolling force prediction device based on IDBO-SA-LSTM applied to the online prediction stage according to an embodiment of this application.

[0124] like Figure 11 As shown, the IDBO-SA-LSTM-based cold rolling force prediction device 20 applied in the online prediction stage includes: an acquisition module 201, a calculation module 202, and a prediction module 203.

[0125] The acquisition module 201 is used to acquire actual data during the stable rolling process of the target cold rolling mill.

[0126] The calculation module 202 is used to normalize and calculate the fitness of the actual data to obtain the corresponding fitness calculation index.

[0127] The prediction module 203 is used to input actual data into the pre-constructed target SA-LSTM rolling force prediction model and combine it with the fitness calculation index to output the rolling force prediction result during the smooth rolling process of the target cold continuous rolling mill. The target SA-LSTM rolling force prediction model is obtained by training an adaptive long short-term memory network optimized by a pre-defined improved dung beetle optimization algorithm on the pre-defined training dataset and test dataset of the cold continuous rolling force.

[0128] It should be noted that the foregoing explanation of the embodiment of the cold continuous rolling force prediction method based on IDBO-SA-LSTM applied to the online prediction stage also applies to the cold continuous rolling force prediction device based on IDBO-SA-LSTM applied to the online prediction stage of this embodiment, and will not be repeated here.

[0129] The cold continuous rolling force prediction device based on IDBO-SA-LSTM proposed in this application, applied to the online prediction stage, includes an acquisition module 201 for acquiring actual data during the smooth rolling process of the target cold continuous rolling mill; a calculation module 202 for normalizing and calculating the fitness of the actual data to obtain the corresponding fitness calculation index; and a prediction module 203 for inputting the actual data into a pre-constructed target SA-LSTM rolling force prediction model and combining it with the fitness calculation index to output the rolling force prediction result during the smooth rolling process of the target cold continuous rolling mill. The target SA-LSTM rolling force prediction model is obtained by training an adaptive long short-term memory network optimized by a pre-defined improved dung beetle optimization algorithm using a pre-defined training dataset and a test dataset of cold continuous rolling force. This application optimizes the long short-term memory network combined with a self-attention mechanism using an improved dung beetle optimization algorithm, effectively improving the accuracy and generalization performance of cold continuous rolling force prediction, and achieving high-precision rolling force prediction for the aluminum foil cold continuous rolling process.

[0130] Figure 12 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include:

[0131] The memory 1201, the processor 1202, and the computer program stored on the memory 1201 and executable on the processor 1202.

[0132] When the processor 1202 executes the program, it implements the cold continuous rolling force prediction method based on IDBO-SA-LSTM provided in the above embodiments.

[0133] Furthermore, electronic devices also include:

[0134] Communication interface 1203 is used for communication between memory 1201 and processor 1202.

[0135] The memory 1201 is used to store computer programs that can run on the processor 1202.

[0136] The memory 1201 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0137] If the memory 1201, processor 1202, and communication interface 1203 are implemented independently, then the communication interface 1203, memory 1201, and processor 1202 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be divided into address buses, data buses, control buses, etc. For ease of representation, Figure 12 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0138] Optionally, in a specific implementation, if the memory 1201, processor 1202, and communication interface 1203 are integrated on a single chip, then the memory 1201, processor 1202, and communication interface 1203 can communicate with each other through an internal interface.

[0139] The processor 1202 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0140] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predicting cold continuous rolling force based on IDBO-SA-LSTM.

[0141] This application also provides a computer program product, including a computer program, which, when executed, is used to implement the above-described method for predicting cold continuous rolling force based on IDBO-SA-LSTM.

[0142] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0143] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0144] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0145] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0146] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0147] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0148] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0149] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A cold continuous rolling force prediction method based on IDBO-SA-LSTM, applied in the offline training stage, characterized in that, Includes the following steps: Based on the preset rolling force calculation model, Circle chaotic mapping strategy, golden sine strategy and dynamic weight coefficient, the preset dung beetle optimization algorithm is optimized to construct an improved dung beetle optimization algorithm. The improved dung beetle optimization algorithm is tested using a variety of preset benchmark test functions to obtain an improved dung beetle optimization algorithm that meets the preset performance test requirements. Based on the improved dung beetle optimization algorithm, the preset adaptive long short-term memory network is automatically optimized, and a target SA-LSTM rolling force prediction model is constructed through the automatically optimized adaptive long short-term memory network, so as to use the target SA-LSTM rolling force prediction model to predict the rolling force in real time during the online prediction stage. The step of automatically optimizing a preset adaptive long short-term memory network based on the improved dung beetle optimization algorithm, and constructing a target SA-LSTM rolling force prediction model through the automatically optimized adaptive long short-term memory network, includes: The adaptive long short-term memory network is constructed based on a preset self-attention mechanism, LSTM layer, activation layer, fully connected layer and regression layer; Obtain raw data of cold continuous rolling force, and perform random shuffling and normalization operations on the raw data of cold continuous rolling force to construct a target training format dataset, and divide the target training format dataset to generate a training dataset and a test dataset. Multiple parameters in the improved dung beetle optimization algorithm are initialized, and the dung beetle population corresponding to the improved dung beetle optimization algorithm is initialized based on the Circle chaotic mapping strategy to obtain multiple individual positions. The adaptive long short-term memory network is trained using the multiple individual positions to generate individual positions with target fitness. The multiple parameters include population size, maximum number of iterations, number of hidden layer neurons, initial learning rate, regularization coefficient, and the proportion of rolling, breeding, foraging, and stealing behaviors in the population. The improved dung beetle optimization algorithm is used to update the individual position of the target fitness, and the iteration is repeated until the maximum number of iterations is reached to obtain the optimal solution parameters. The adaptive long short-term memory network is then trained using the optimal solution parameters and the training dataset to obtain the initial SA-LSTM rolling force prediction model. The initial SA-LSTM rolling force prediction model is fine-tuned using the test dataset to obtain the target SA-LSTM rolling force prediction model.

2. The method for predicting rolling force in cold continuous rolling mills based on IDBO-SA-LSTM according to claim 1, characterized in that, The step of performing performance tests on the improved dung beetle optimization algorithm using multiple preset benchmark test functions to obtain an improved dung beetle optimization algorithm that meets preset performance test requirements includes: The plurality of benchmark testing functions are determined, and the population size and maximum number of iterations for each of the plurality of benchmark testing functions are set for each single iteration. The plurality of benchmark testing functions include a plurality of single-peak testing functions and a plurality of multi-peak testing functions. The improved dung beetle optimization algorithm is tested using each of the test functions based on the population size during a single iteration and the maximum number of iterations.

3. A cold continuous rolling force prediction method based on IDBO-SA-LSTM, applied to the online prediction stage, characterized in that, The method for predicting cold continuous rolling force based on IDBO-SA-LSTM for offline training, as described in any one of claims 1-2, comprises the following steps: Obtain actual data during the stable rolling process of the target cold continuous rolling mill; The actual data is normalized and fitness is calculated to obtain the corresponding fitness calculation index; The actual data is input into the pre-constructed target SA-LSTM rolling force prediction model, and combined with the fitness calculation index, to output the rolling force prediction result during the smooth rolling process of the target cold continuous rolling mill. The target SA-LSTM rolling force prediction model is obtained by training an adaptive long short-term memory network optimized by a pre-defined improved dung beetle optimization algorithm on a pre-defined training dataset and test dataset of cold continuous rolling force.

4. A cold continuous rolling force prediction device based on IDBO-SA-LSTM, applied in the offline training stage, characterized in that, include: The optimization module is used to optimize the preset dung beetle optimization algorithm based on the preset rolling force calculation model, Circle chaotic mapping strategy, golden sine strategy and dynamic weight coefficients, so as to construct an improved dung beetle optimization algorithm. The testing module is used to perform performance tests on the improved dung beetle optimization algorithm through a variety of preset benchmark testing functions, so as to obtain an improved dung beetle optimization algorithm that meets the preset performance testing requirements. The modeling module is used to automatically optimize the preset adaptive long short-term memory network based on the improved dung beetle optimization algorithm, and to construct a target SA-LSTM rolling force prediction model through the automatically optimized adaptive long short-term memory network, so as to use the target SA-LSTM rolling force prediction model to predict the rolling force in real time during the online prediction stage. The modeling module includes: The building unit is used to construct the adaptive long short-term memory network based on a preset self-attention mechanism, LSTM layer, activation layer, fully connected layer and regression layer; A partitioning unit is used to acquire raw data of cold rolling force, and to perform random shuffling and normalization operations on the raw data of cold rolling force to construct a target training format dataset, and to partition the target training format dataset to generate a training dataset and a test dataset. An initialization unit is used to initialize multiple parameters in the improved dung beetle optimization algorithm and, based on the Circle chaotic mapping strategy, initialize the dung beetle population corresponding to the improved dung beetle optimization algorithm to obtain multiple individual positions. The adaptive long short-term memory network is then trained using the multiple individual positions to generate individual positions with target fitness. The multiple parameters include the population size, maximum number of iterations, number of hidden layer neurons, initial learning rate, regularization coefficient, and the proportion of behaviors such as rolling, breeding, foraging, and stealing in the population. An iterative unit is used to update the individual position of the target fitness using the improved dung beetle optimization algorithm and iterates to the maximum number of iterations to obtain the optimal solution parameters. The adaptive long short-term memory network is then trained using the optimal solution parameters and the training dataset to obtain the initial SA-LSTM rolling force prediction model. The fine-tuning unit is used to fine-tune the initial SA-LSTM rolling force prediction model using the test dataset to obtain the target SA-LSTM rolling force prediction model.

5. The cold continuous rolling force prediction device based on IDBO-SA-LSTM according to claim 4, characterized in that, The modeling module includes: The building unit is used to construct the adaptive long short-term memory network based on a preset self-attention mechanism, LSTM layer, activation layer, fully connected layer and regression layer; A partitioning unit is used to acquire raw data of cold rolling force, and to perform random shuffling and normalization operations on the raw data of cold rolling force to construct a target training format dataset, and to partition the target training format dataset to generate a training dataset and a test dataset. An initialization unit is used to initialize multiple parameters in the improved dung beetle optimization algorithm and, based on the Circle chaotic mapping strategy, initialize the dung beetle population corresponding to the improved dung beetle optimization algorithm to obtain multiple individual positions. The adaptive long short-term memory network is then trained using the multiple individual positions to generate individual positions with target fitness. The multiple parameters include the population size, maximum number of iterations, number of hidden layer neurons, initial learning rate, regularization coefficient, and the proportion of behaviors such as rolling, breeding, foraging, and stealing in the population. An iterative unit is used to update the individual position of the target fitness using the improved dung beetle optimization algorithm and iterates to the maximum number of iterations to obtain the optimal solution parameters. The adaptive long short-term memory network is then trained using the optimal solution parameters and the training dataset to obtain the initial SA-LSTM rolling force prediction model. The fine-tuning unit is used to fine-tune the initial SA-LSTM rolling force prediction model using the test dataset to obtain the target SA-LSTM rolling force prediction model.

6. A cold continuous rolling force prediction device based on IDBO-SA-LSTM, applied in the online prediction stage, characterized in that, The IDBO-SA-LSTM-based cold continuous rolling force prediction device for offline training phase, as described in any one of claims 4-5, comprises: The acquisition module is used to acquire actual data during the smooth rolling process of the target cold continuous rolling mill; The calculation module is used to normalize and calculate the fitness of the actual data to obtain the corresponding fitness calculation index. The prediction module is used to input the actual data into the pre-constructed target SA-LSTM rolling force prediction model and combine it with the fitness calculation index to output the rolling force prediction result during the smooth rolling process of the target cold rolling mill. The target SA-LSTM rolling force prediction model is obtained by training an adaptive long short-term memory network optimized by a pre-defined improved dung beetle optimization algorithm on a pre-defined training dataset and test dataset of cold rolling force.

7. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, the processor executing the program to implement the cold continuous rolling force prediction method based on IDBO-SA-LSTM as described in any one of claims 1-2 or 3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the cold continuous rolling force prediction method based on IDBO-SA-LSTM as described in any one of claims 1-2 or 3.

9. A computer program product, comprising a computer program, characterized in that, The computer program is executed by a processor to implement the cold continuous rolling force prediction method based on IDBO-SA-LSTM as described in any one of claims 1-2 or 3.

Images