A single set of header cooling flow prediction method based on ETR model

Abstract

The present application discloses a method for predicting the cooling flow of a single group of headers based on the ETR model, which belongs to the field of hot rolling cooling technology. The method comprises: collecting post-rolling cooling data in real time and extracting characteristic data related to the cooling flow of a single group of headers, establishing a database to train the ETR model after pre-processing, determining the number of optimal trees, establishing an optimization model through grid search hyperparameter combination, selecting appropriate indicators to evaluate the applicability of the optimization model according to actual application conditions, and dynamically guiding the cooling process procedures in real time. The prediction method of the present application has a high degree of automation, which can not only realize the rapid and accurate prediction of the cooling flow of a single group of headers, but also shorten the response time of the overall cooling system, effectively save production costs, and has good prospects for industrial application.

Description

Technical Field

[0001] The present invention relates to the field of hot rolling cooling technology, and in particular to a method for predicting the cooling flow of a single group of headers based on an ETR model. Background Art

[0002] From ancient times to the present, steel products have been an indispensable raw material for the national economy. With the development of the economy and society, market expectations for the specifications and quality of steel products have also increased. Companies are also gradually shifting their production models from large-scale, single-source approaches to diversified and diverse development. In this context, complex operating conditions and ever-changing product specifications have brought unprecedented challenges to traditional cooling systems, especially in terms of their adaptability and adjustment capabilities. The market has placed more stringent demands on these systems to better adapt and meet current needs.

[0003] By controlling the cooling flow of a single set of headers, flexible control of cooling systems of different specifications can be achieved. In actual production applications, a 20-point automatic calibration method is often used for control. That is, the corresponding functional relationship between the opening degree of the regulating valve of each header and the flow rate is obtained by interpolation based on the 20-point flow curve, and the flow rate of each injection header is controlled by the L1 system. However, due to differences in factors such as the boiling water method and water volume for different specifications of steel grades on site, the effect of adjusting the cooling flow rate based on this method is poor. Additional closed-loop adjustments are still required during the actual boiling water process, which increases the flow stabilization time and water consumption costs. In recent years, the vigorous development of the field of artificial intelligence has provided new ideas for the control of single-set header flow. It is urgent to select appropriate algorithms and training models to obtain a highly automated single-set header cooling flow prediction method with industrial application value. Summary of the Invention

[0004] In view of this, the present application provides a single-group header cooling flow prediction method based on the ETR model, which mainly aims to solve the problems of low automation level, low reliability of prediction results, low prediction efficiency and high energy consumption in the current single-group header cooling flow prediction method.

[0005] To achieve the above objectives, this application provides the following technical solutions, including:

[0006] Taking the hot rolling cooling system as the research object, the L1 to L3 systems dynamically and in real time collect complete post-rolling cooling data. Characteristic data related to the cooling flow rate of a single group of headers is extracted from the complete cooling data, focusing on the process, equipment, and energy consumption. Based on the properties and application conditions of the characteristic data set, it is preprocessed to ensure the accuracy of the model's characteristic variables. Based on the preprocessed characteristic data, a header cooling flow control database consisting of training and test sets is constructed.

[0007] The ETR model is trained using the training set, the number of optimal trees in the ETR algorithm is determined using a mean square error metric, and appropriate hyperparameters, an algorithm for retrieving hyperparameter combinations, and a method for evaluating the performance of hyperparameter combinations are selected based on the properties of a header cooling flow control database and application conditions, thereby establishing an optimized ETR model for predicting a single group of header cooling flow rates;

[0008] The optimized ETR model is tested using the test set, and appropriate model evaluation indicators are selected according to its application conditions to judge the accuracy of the optimized ETR model. The applicability of the optimized ETR model in predicting the cooling flow conditions of a single set of headers is comprehensively verified in combination with the actual response time and energy consumption, providing real-time dynamic guidance for the cooling process procedures.

[0009] According to the above technical solution, the present application provides a single-group header cooling flow prediction method based on the ETR model with a high degree of automation. It can dynamically optimize the prediction results in real time according to production data to guide the cooling process procedures. Compared with traditional prediction methods, it effectively reduces the shutdown adjustment response time in the actual water boiling process, thereby improving production efficiency while reducing water consumption and saving production costs.

[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] By reading the detailed description of the preferred embodiment below, various other advantages and benefits will become clear to those skilled in the art. The accompanying drawings are only used to illustrate the preferred embodiment and are not to be considered as limiting the present application, wherein:

[0012] Figure 1 : is a graph showing the relationship between the number of trees used in the model according to the embodiment of the present application and the corresponding mean square error, where the horizontal axis is the number of trees and the vertical axis is the mean square error;

[0013] Figure 2 It is a flow chart of a method for predicting cooling flow of a single group of headers provided in an embodiment of the present application. DETAILED DESCRIPTION

[0014] The following describes the embodiments of the present invention through specific examples. Those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. Among them, the accompanying drawings are only for illustrative purposes and only represent schematic diagrams, not physical drawings, and cannot be understood as limiting the present application. In order to better illustrate the embodiments of the present application, it is understandable to those skilled in the art that some well-known structures or steps in the drawings and their descriptions may be omitted.

[0015] Reference Figure 1 and Figure 2 Taking a hot rolling production line of a certain factory as an example, a single group header cooling flow prediction method based on the ETR model involved in the implementation of this application is explained.

[0016] The present application provides a method for predicting cooling flow of a single set of headers based on an ETR model, including:

[0017] Taking the hot rolling cooling system as the research object, complete post-rolling cooling data was collected dynamically and in real time through the L1 to L3 systems. Considering that the manifold equipment was in a maintained state, characteristic data related to the cooling flow of a single manifold was extracted from the complete cooling data, including the booster water supply pump outlet manifold pressure, valve opening, number of cooling manifold groups opened, actual cooling flow of a single manifold, and flow stabilization time. A total of 516 raw data items were obtained. To ensure the accuracy of the model's characteristic variables, the raw data were subjected to data augmentation processing by interpolation. The generated new data were merged with the original data to expand the data volume to 1000 items. This helped improve the generalization of the single manifold cooling flow prediction model and reduce the risk of overfitting. The preprocessed characteristic data were divided into training and test sets with an 8:2 ratio to construct a manifold cooling flow control database. The valve opening, booster water supply pump outlet manifold pressure, and number of cooling manifold groups opened were used as influencing factors, and the actual cooling flow of a single manifold was stored as the result in the established manifold cooling flow control database.

[0018] The ETR model is trained using the training set, such as Figure 1 As shown, the following formula is used to calculate the mean square error corresponding to each tree and draw a graph showing the relationship between the number of trees used in the model and the mean square error.

[0019]

[0020] Where: MSE is the mean square error, n is the number of samples; y i is the true value of the sample; is the sample prediction value. When the mean square error is minimized, the number of optimal trees in the ETR model can be obtained, such as Figure 1 As shown, the number of optimal trees this time is n=131.

[0021] Considering that the single-group header flow prediction is a small-dimensional sample, the selected hyperparameters include the maximum depth, the minimum number of samples for node partitioning, and the minimum number of samples for leaf nodes. Specifically, the search range of the maximum depth hyperparameter is (None, 10, 20), the search range of the minimum number of samples for node partitioning is (2, 5, 10), and the search range of the minimum number of samples for leaf nodes is (1, 2, 4). In addition, the ETR algorithm that uses grid search for hyperparameter optimization has a simple structure and a short running time in the case of small sample data. Compared with random search and Bayesian optimization, it is more suitable for application in the prediction scenario of a single group of header flow, and can complete the prediction of a single group of header cooling flow in a short time. Therefore, it is determined to use the grid search algorithm to try different combinations within a given hyperparameter range, and use k-fold cross-validation to evaluate the performance of the above hyperparameter combinations. According to the performance scores of each combination, the algorithm model selects the combination with the best performance index as the optimal model configuration. After using grid search for hyperparameter optimization and using cross-validation to evaluate the model performance, this embodiment finally determines to establish an optimization model for predicting a single group of header cooling flow with a maximum hyperparameter depth of 20, a minimum number of samples for node division of 20, and a minimum number of samples for leaf nodes of 1.

[0022] Similarly, considering that the samples for predicting the flow rate of a single group of headers are small and of small dimensions, the model evaluation indicators selected include root mean square error, average error, and determination coefficient. Then, the applicability of the self-learning method is analyzed by comparing the predicted results with the actual results. As shown in Table 1, the optimized ETR model was tested on the test set. The results show that the root mean square error of the ETR model optimized by grid search hyperparameters in this embodiment is 4.5920, the average error is 2.9364, and the determination coefficient is 0.9824. Compared with the random forest and default parameter ETR models, the ETR model optimized by grid search hyperparameters has lower root mean square error and average error, and the determination coefficient is closer to 1. This shows that the ETR model optimized by grid search hyperparameters is more suitable for predicting the cooling flow rate of a single group of headers. It has higher accuracy and can also guide the cooling process procedures in real time according to the dynamically updated post-rolling cooling data, with a high degree of automation.

[0023] Table 1 Prediction performance results of each model

[0024]

[0025] The optimized ETR model described above was applied to predict the cooling flow rate of a single manifold set on an actual production line. The results showed that, compared to the ~7s response time achieved with a 20-point calibration, the actual response time of the optimized ETR model provided in this application was reduced to ~1 / 2 of the original, to ~3s. This increased production efficiency directly reduced water and electricity consumption, saving production costs. Furthermore, based on actual post-rolling cooling data collected, the optimized ETR model's predictions for cooling flow rates of a single manifold set were slightly more accurate than those achieved with the 20-point calibration.

[0026] The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatus.

[0027] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for predicting cooling flow of a single set of headers based on the ETR model, characterized in that: include: Taking the hot rolling cooling system as the research object, the L1 to L3 systems dynamically and in real time collect complete post-rolling cooling data. Characteristic data related to the cooling flow of a single header group is extracted from the complete cooling data based on the process, equipment, and energy consumption. Based on the properties and application conditions of the characteristic data, it is preprocessed to ensure the accuracy of the model characteristic variables. Based on the preprocessed characteristic data, a header cooling flow control database containing training and test sets is constructed. The relevant characteristic data includes the outlet main pressure of the booster water supply pump, valve opening degree, the number of cooling header groups open, the actual cooling flow of a single header group, the status of the header equipment, and the flow stabilization time. The ETR model is trained using the training set, the number of optimal trees in the ETR algorithm is determined using a mean square error metric, and appropriate hyperparameters, an algorithm for retrieving hyperparameter combinations, and a method for evaluating the performance of hyperparameter combinations are selected based on the properties of a header cooling flow control database and application conditions, thereby establishing an optimized ETR model for predicting a single group of header cooling flow rates; The optimized ETR model is tested using the test set, and appropriate model evaluation indicators are selected according to its application conditions to judge the accuracy of the optimized ETR model. The applicability of the optimized ETR model in predicting the cooling flow conditions of a single set of headers is comprehensively verified in combination with the actual response time and energy consumption, providing real-time dynamic guidance for the cooling process procedures.

2. The method for predicting cooling flow of a single set of headers based on the ETR model according to claim 1, characterized in that: In order to improve the generalization of the cooling flow prediction model of a single group of headers and reduce the risk of overfitting, the preprocessing is a data enhancement process of interpolating the feature data.

3. The method for predicting cooling flow of a single set of headers based on the ETR model according to claim 1, characterized in that: The hyperparameters include the maximum depth, the minimum number of samples for node partitioning, and the minimum number of samples for leaf nodes. The algorithm for retrieving hyperparameter combinations is a grid search algorithm, and the method for evaluating the performance of hyperparameter combinations is a k-fold cross-validation method.

4. The method for predicting cooling flow of a single set of headers based on the ETR model according to claim 1, characterized in that: The model evaluation indicators include root mean square error, mean error and determination coefficient.

Images