Seamless steel tube outer diameter prediction method and device based on mixed kernel function SVR model

Abstract

The invention provides a seamless steel tube outer diameter prediction method and device based on a mixed kernel function SVR model, and relates to the technical field of seamless steel tube rolling. The method comprises the following steps: selecting preliminary characteristics influencing the outer diameter of the seamless steel pipe, and obtaining related data for preprocessing; calculating the importance degree of each preliminary feature by adopting a random forest model, and selecting important features to generate training data; and constructing a data-driven seamless steel tube outer diameter prediction model by adopting an SVR model based on a mixed kernel function, and obtaining a seamless steel tube outer diameter prediction result according to the trained seamless steel tube outer diameter prediction model. According to the method, the SVR model is constructed in a mixed kernel function mode, so that the local learning ability and the global generalization ability of the model are balanced, and the outer diameter prediction precision is improved. Meanwhile, through the outer diameter prediction model, the outer diameter of a finished pipe can be predicted in the parameter setting stage, the process parameters are corrected in advance to reduce outer diameter errors, and the influence caused by control lag of the process parameters is eliminated.

Description

Technical Field

[0001] The invention relates to the technical field of seamless steel pipe rolling, and in particular to a method and a device for predicting the outer diameter of a seamless steel pipe based on a mixed kernel function SVR model. Background Art

[0002] The outer diameter has always been one of the important indicators of the quality of seamless steel pipes. At the same time, customers have increasingly higher requirements for the outer diameter control accuracy of seamless steel pipes. Therefore, improving the outer diameter control accuracy of seamless steel pipes has become an important task to improve the quality of seamless steel pipe products. However, during on-site production, the control of outer diameter accuracy is still based on the production experience of on-site technicians. According to the size of the previous steel pipe, the production process parameters of each process section are manually modified, which leads to the lag and uncertainty of process parameter control during on-site production. The outer diameter deviation of the first few steel pipes in most batches is large, and one or two process parameter modifications are required to better ensure the outer diameter control accuracy, which increases the probability of the first few steel pipes being scrapped due to outer diameter deviation, resulting in production waste. Therefore, it is necessary to establish a prediction model based on the relationship between the outer diameter of the steel pipe and the process parameters of each deformation stage, predict the outer diameter of the finished pipe in the parameter setting stage, reduce the outer diameter error by correcting the process parameters in advance, and eliminate the impact caused by the lag in process parameter control.

[0003] The hot rolling production process of seamless steel pipes involves three major deformation processes: perforation, tube rolling, and diameter reduction. The parameter changes of these processes have a complex and significant impact on the outer diameter of the steel pipe, showing multivariate, strong coupling and nonlinear characteristics. The traditional analytical model cannot accurately calculate the outer diameter of the steel pipe. Therefore, it is necessary to use artificial intelligence methods driven by industrial data to predict the outer diameter of seamless steel pipes and improve its prediction accuracy, thereby improving the control accuracy of the production site.

[0004] In the traditional process of seamless steel pipe outer diameter prediction, people often only focus on the dimensional accuracy prediction of one process, without considering the impact of multiple processes on the dimensional accuracy of seamless steel pipes, and ignore the impact of roll wear and thermal deformation in the actual production process, resulting in a large error range in the prediction of the seamless steel pipe outer diameter and low model prediction accuracy. Summary of the invention

[0005] In order to solve the technical problems of hysteresis and uncertainty of process parameter control during on-site production and low prediction accuracy of traditional prediction models in the prior art, an embodiment of the present invention provides a method and device for predicting the outer diameter of a seamless steel pipe based on a hybrid kernel function SVR model. The technical solution is as follows:

[0006] On the one hand, a method for predicting the outer diameter of a seamless steel pipe based on a hybrid kernel function SVR model is provided, the method being implemented by a seamless steel pipe outer diameter prediction device, the method comprising:

[0007] S1. Select the preliminary features that affect the outer diameter of seamless steel pipes, obtain the relevant data of the preliminary features, preprocess the relevant data, and obtain the preprocessed data.

[0008] S2. According to the preprocessed data, use the random forest model to calculate the importance of each preliminary feature, select the important features according to the importance, and generate training data based on the important features.

[0009] S3. Construct a data-driven prediction model for the outer diameter of seamless steel pipes using the SVR model based on a hybrid kernel function, train the prediction model for the outer diameter of seamless steel pipes according to the training data, and obtain a trained prediction model for the outer diameter of seamless steel pipes.

[0010] S4. Obtain the data of the seamless steel pipes to be predicted, input them into the trained prediction model for the outer diameter of seamless steel pipes, and obtain the prediction results of the outer diameter of seamless steel pipes.

[0011] Optionally, the selection of the preliminary features that affect the outer diameter of seamless steel pipes in S1 includes:

[0012] Select the process parameters that affect the outer diameter of seamless steel pipes based on the mechanism knowledge of the piercing stage, rolling stage, and reducing stage.

[0013] Take the process parameters, downtime, and historical rolling length as preliminary features.

[0014] Among them, the process parameters include the tapping temperature, roll gap, plug diameter, plug length, extension, feeding angle, rolling angle, guide distance, roll speed, and roll biting speed in the piercing stage; the throat diameter, mandrel diameter, mandrel speed, feeding angle, rolling angle, and roll speed in the rolling stage; the last stand pass, number of stands, roll speed of each stand, and data of the incoming steel grade and incoming specification in the reducing stage.

[0015] Optionally, the downtime and historical rolling length are used to consider the influence of roll thermal deformation and roll wear on the outer diameter of steel pipes during the rolling process.

[0016] The process of obtaining the downtime includes:

[0017] Taking the time point when the head of the steel pipe enters the preset position as the reference, calculate the production interval time between each steel pipe and the previous steel pipe, and obtain the downtime of each steel pipe.

[0018] The process of obtaining the historical rolling length includes:

[0019] Calibrate the historical rolling length of the first steel pipe produced in the collected data to 0, and the historical rolling length of each subsequent steel pipe is equal to the historical rolling length of the previous steel pipe plus the billet length of the previous steel pipe, so as to obtain the historical rolling length of each steel pipe.

[0020] Optionally, for the relevant data for obtaining preliminary features in S1, preprocess the relevant data to obtain preprocessed data, including:

[0021] S11. According to the time recorded by the branch-by-branch tracking system, intercept the process parameter curve of each steel pipe; take the mean or maximum value of the process parameter curve to generate production process parameter data; match the tooling data, basic information, and production process parameter data according to the steel pipe number, and generate relevant data of preliminary features based on the time recorded by the diameter gauge.

[0022] S12. Perform data transformation processing and standardization processing on the relevant data to obtain preprocessed data.

[0023] Optionally, in S2, use a random forest model to calculate the importance degree of each preliminary feature, select important features according to the importance degree, and generate training data based on the important features, including:

[0024] S21. Use the preprocessed data to train the random forest model, and calculate the importance of each preliminary feature according to the trained forest model.

[0025] S22. Select a certain preliminary feature, and randomly rearrange all the feature values of the selected preliminary feature.

[0026] S23. Calculate the importance of the rearranged preliminary feature according to the trained forest model.

[0027] S24. Calculate the importance score of the preliminary feature by comparing the difference between the importance of the preliminary feature and the importance of the rearranged preliminary feature.

[0028] S25. Determine whether each preliminary feature has been selected. If so, generate training data according to the importance score of the preliminary feature; if not, go to step S22.

[0029] Optionally, the calculation of the importance score of the preliminary feature in S24 is shown in the following formula (1):

[0030] (1)

[0031] In the formula, represents the importance score of the preliminary feature, represents the importance of the preliminary feature, represents the importance of the rearranged preliminary feature.

[0032] Optionally, in S3, use an SVR model based on a hybrid kernel function to construct a data-driven seamless steel pipe outer diameter prediction model, and train the seamless steel pipe outer diameter prediction model according to the training data, including:

[0033] S31. Construct a data-driven seamless steel pipe outer diameter prediction model using the SVR model based on the hybrid kernel function.

[0034] Among them, the hybrid kernel function is based on the Poly kernel function and the RBF kernel function, and is constructed according to the following formula (2):

[0035] (2)

[0036] In the formula, represents the hybrid kernel function, represents the mixing coefficient, represents the RBF kernel function expression, represents the Poly kernel function expression.

[0037] S32. Train the seamless steel pipe outer diameter prediction model according to the training data, and use the PSO algorithm to optimize the hyperparameters of the SVR model based on the hybrid kernel function.

[0038] Among them, the hyperparameters include: regularization coefficient, RBF kernel function coefficient, misclassification tolerance, and mixing coefficient.

[0039] On the other hand, a seamless steel pipe outer diameter prediction device based on the SVR model with a hybrid kernel function is provided. This device is applied to the seamless steel pipe outer diameter prediction method based on the SVR model with a hybrid kernel function. The device includes:

[0040] An acquisition model, which is used to select the preliminary features that affect the outer diameter of the seamless steel pipe, acquire the relevant data of the preliminary features, preprocess the relevant data, and obtain the preprocessed data.

[0041] A generation model, which is used to calculate the importance degree of each preliminary feature according to the preprocessed data using the random forest model, select the important features according to the importance degree, and generate training data according to the important features.

[0042] A training model, which is used to construct a data-driven seamless steel pipe outer diameter prediction model using the SVR model based on the hybrid kernel function, train the seamless steel pipe outer diameter prediction model according to the training data, and obtain the trained seamless steel pipe outer diameter prediction model.

[0043] An output model, which is used to obtain the seamless steel pipe data to be predicted, input it into the trained seamless steel pipe outer diameter prediction model, and obtain the seamless steel pipe outer diameter prediction result.

[0044] Optionally, the acquisition model is further used for:

[0045] Select the process parameters that affect the outer diameter of the seamless steel pipe based on the mechanism knowledge of the piercing stage, rolling pipe stage, and reducing diameter stage.

[0046] Take process parameters, downtime, and historical rolling length as preliminary features.

[0047] Among them, the process parameters include the furnace outlet temperature, roll gap, plug diameter, plug length, elongation, feeding angle, rolling angle, guide distance, roll speed, and roll biting speed during the piercing stage; the throat diameter, mandrel diameter, mandrel speed, feeding angle, rolling angle during the pipe rolling stage, and roll speed; the pass of the last stand, the number of stands, the roll speed of each stand, and the incoming steel grade and incoming specification data during the reducing stage.

[0048] Optionally, the downtime and historical rolling length are used to consider the influence of roll thermal deformation and roll wear on the outer diameter of the steel pipe during the rolling process.

[0049] The process of obtaining the downtime includes:

[0050] Based on the time point when the head of the steel pipe enters the preset position, calculate the production interval time between each steel pipe and the previous one to obtain the downtime of each steel pipe.

[0051] The process of obtaining the historical rolling length includes:

[0052] Calibrate the historical rolling length of the first steel pipe produced in the collected data to 0, and then the historical rolling length of each subsequent steel pipe is equal to the historical rolling length of the previous steel pipe plus the billet length of the previous steel pipe to obtain the historical rolling length of each steel pipe.

[0053] Optionally, the obtained model is further used for:

[0054] S11. According to the time recorded by the piece-by-piece tracking system, intercept the process parameter curve of each steel pipe; take the mean or maximum value of the process parameter curve to generate production process parameter data; match the tooling data, basic information, and production process parameter data according to the steel pipe number, and generate relevant data of the preliminary features based on the time recorded by the diameter gauge.

[0055] S12. Perform data transformation processing and standardization processing on the relevant data to obtain the preprocessed data.

[0056] Optionally, the generated model is further used for:

[0057] S21. Use the preprocessed data to train a random forest model, and calculate the importance of each preliminary feature according to the trained forest model.

[0058] S22. Select a certain preliminary feature and randomly rearrange all the feature values of the selected preliminary feature.

[0059] S23. Calculate the importance of the rearranged preliminary feature according to the trained forest model.

[0060] S24. Calculate the importance score of the preliminary features by comparing the difference between the importance of the preliminary features and the importance of the rearranged preliminary features.

[0061] S25. Determine whether each preliminary feature is selected. If so, generate training data based on the importance scores of the preliminary features; if not, go back to step S22.

[0062] Optionally, the calculation of the importance score of the preliminary features in S24 is shown in the following formula (1):

[0063] (1)

[0064] In the formula, represents the importance score of the preliminary features, represents the importance of the preliminary features, represents the importance of the rearranged preliminary features.

[0065] Optionally, the trained model is further used for:

[0066] S31. Construct a data-driven seamless steel pipe outer diameter prediction model using an SVR model based on a hybrid kernel function.

[0067] Among them, the hybrid kernel function is based on the Poly kernel function and the RBF kernel function as the basic kernel functions, and is constructed according to the following formula (2):

[0068] (2)

[0069] In the formula, represents the hybrid kernel function, represents the mixing coefficient, represents the RBF kernel function expression, represents the Poly kernel function expression.

[0070] S32. Train the seamless steel pipe outer diameter prediction model according to the training data, and use the PSO algorithm to optimize the hyperparameters of the SVR model based on the hybrid kernel function.

[0071] Among them, the hyperparameters include: regularization coefficient, RBF kernel function coefficient, misclassification tolerance, and mixing coefficient.

[0072] On the other hand, a seamless steel pipe outer diameter prediction device is provided. The seamless steel pipe outer diameter prediction device includes: a processor; a memory, and computer-readable instructions are stored on the memory. When the computer-readable instructions are executed by the processor, any one of the methods in the above seamless steel pipe outer diameter prediction method based on the hybrid kernel function SVR model is implemented.

[0073] On the other hand, a computer-readable storage medium is provided, in which at least one instruction is stored, and the at least one instruction is loaded and executed by a processor to implement any one of the above seamless steel pipe outer diameter prediction methods based on the hybrid kernel function SVR model.

[0074] The beneficial effects brought by the technical solutions provided in the embodiments of the present invention at least include:

[0075] In the embodiments of the present invention, considering the comprehensive influence of the process parameters in the three major deformation stages of piercing, rolling, and sizing on the outer diameter of the steel pipe, the preliminary screening of characteristic variables is carried out based on the mechanism knowledge of hot rolling of steel pipes and data analysis, and the influence of roll thermal deformation and roll wear during the rolling process is considered by adding the downtime and historical rolling length. Finally, through the random forest model using the idea of permutation importance, the importance degree of each characteristic variable is calculated, redundant variables are removed, and a high-quality dataset is obtained, which helps to improve the accuracy of the seamless steel pipe outer diameter prediction model.

[0076] The present invention optimizes the SVR model by using a method of hybrid Poly kernel function and RBF kernel function, which is beneficial to improving the prediction accuracy and generalization ability of the model.

[0077] The present invention predicts the outer diameter of the finished pipe according to the process parameters of the three deformation processes of piercing, rolling, and sizing, which can be used for the diagnosis of the outer diameter deviation of hot-rolled seamless steel pipes and the optimization of process parameters, and is beneficial to eliminating the lag of process parameter adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0079] Figure 1 is a flowchart of a seamless steel pipe outer diameter prediction method based on a hybrid kernel function SVR model provided by an embodiment of the present invention;

[0080] Figure 2 is a comparison chart of the prediction accuracy of the hybrid kernel function SVR model considering "downtime" and "historical rolling length" and not considering "downtime" and "historical rolling length" provided by an embodiment of the present invention;

[0081] Figure 3 is a ranking chart of the importance degree of each characteristic variable provided by an embodiment of the present invention;

[0082] Figure 4 is a comparison chart of the model prediction deviation distribution provided by an embodiment of the present invention;

[0083] Figure 5 It is a block diagram of a seamless steel pipe outer diameter prediction device based on a hybrid kernel function SVR model provided by an embodiment of the present invention;

[0084] Figure 6 It is a structural schematic diagram of a seamless steel pipe outer diameter prediction device provided by an embodiment of the present invention. Specific embodiments

[0085] The following describes the technical solutions in the present invention with reference to the accompanying drawings.

[0086] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0087] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "Of", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0088] In the embodiments of the present invention, sometimes subscripts such as W 1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0089] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0090] An embodiment of the present invention provides a method for predicting the outer diameter of seamless steel pipes based on a hybrid kernel function SVR model. This method can be implemented by a seamless steel pipe outer diameter prediction device, which can be a terminal or a server. As Figure 1 shown in the flowchart of the method for predicting the outer diameter of seamless steel pipes based on a hybrid kernel function SVR model, the processing flow of this method can include the following steps:

[0091] S1. Select preliminary features that affect the outer diameter of seamless steel pipes, obtain relevant data of the preliminary features, and preprocess the relevant data to obtain preprocessed data.

[0092] Optionally, the preliminary features that the selection in S1 affects the outer diameter of seamless steel pipes may include the following steps A11 - A12:

[0093] A11. Select process parameters that have a greater impact on the outer diameter of the finished pipe as characteristic variables based on the mechanism knowledge of the piercing, rolling, and reducing processes. The preliminarily screened features are shown in Table 1. In the field of seamless steel pipe quality prediction, there is less research on outer diameter prediction. Most research only uses the process parameters of this process to predict the outer diameter or wall thickness at the outlet of a single process. For example, in the prediction of the shape of skew-rolled piercing tubes based on the GRA-based PSO-BP neural network, only 5 quantities, namely the forward feed, roll spacing, guide plate spacing, billet diameter, and plug diameter, are used as input features to predict the outer diameter and wall thickness of the pierced mandrel tube. Each link in the production process of seamless steel pipes affects the quality of the final finished pipe. Predicting only the quality of a single link cannot actually reflect the influence of each link on the quality of the final finished pipe. In addition, most seamless steel pipe quality predictions do not consider the influence of roll wear and roll thermal deformation during the production process. The present invention selects process parameters that have a greater impact on the outer diameter of the finished pipe in the piercing, rolling, and reducing processes as the input of the model, so as to more comprehensively consider the strong coupling and non-linear influence of the piercing, rolling, and reducing processes on the outer diameter of the finished pipe, which is beneficial to improving the prediction accuracy of the outer diameter. At the same time, the features "downtime" and "historical rolling length" are introduced to consider the influence of roll thermal deformation and roll wear on the outer diameter of the steel pipe during the production process. Compared with not considering roll thermal deformation and roll wear, the RMSE and MAE of the hybrid kernel function SVR model are significantly reduced, and the R2 value is significantly improved, as Figure 2 shown.

[0094] Table 1

[0095]

[0096] Among them, the process parameters may include: process parameters in the piercing stage: tapping temperature, roll gap, plug diameter, plug length, forward feed, feeding angle, rolling angle, guide plate distance, roll speed, roll biting speed; process parameters in the rolling stage: throat diameter, mandrel diameter, mandrel speed, feeding angle, rolling angle, roll speed; process parameters in the reducing stage: pass profile of the last stand, number of stands, roll speeds of each stand. In addition, it also includes the incoming steel grade and incoming specification data.

[0097] A12. Use the downtime and historical rolling length as input features to consider the influence of roll thermal deformation and roll wear on the outer diameter of the steel pipe during the rolling process.

[0098] The acquisition process of the downtime includes:

[0099] Taking the time point when the head of the steel pipe enters a certain fixed position as a reference, calculate the production interval time between each steel pipe and the previous one to obtain the downtime data of each steel pipe.

[0100] The process of obtaining the historical rolling length includes:

[0101] Calibrate the historical rolling length of the first steel pipe produced in the collected data to 0. After that, the historical rolling length of each subsequent steel pipe is equal to the historical rolling length of the previous steel pipe plus the billet length of the previous steel pipe to obtain the historical rolling length data of each steel pipe.

[0102] Optionally, for the relevant data of obtaining the preliminary features in S1, preprocess the relevant data to obtain the preprocessed data, including:

[0103] S11. According to the time recorded by the piece-by-piece tracking system, intercept the process parameter curve of each steel pipe, generate production process parameter data by taking the mean or maximum value, and match the tooling data, basic information, and process parameters according to the steel pipe number. Finally, based on the time recorded by the diameter gauge, match the outer diameter of the steel pipe with the relevant data to form an initial training sample.

[0104] S12. Perform data transformation processing and standardization processing on the initial sample.

[0105] In a feasible implementation manner, perform data transformation processing and standardization processing on the initial sample, specifically including:

[0106] S121. Convert the non-numerical data in the initial sample into numerical data.

[0107] Specifically, it may include: for the pass profile feature at the last stand, its format is "-A2" or "11A". Removing the letter A becomes a pure number. The larger the number, the smaller the pass. That is, the number is positively correlated with the pass diameter and is uniquely corresponding to each other.

[0108] Furthermore, for the steel grade feature, match the content of each steel grade C, Si, Mn, Cr, and Ce to the data of each steel pipe one by one to characterize the influence of different steel grades on the outer diameter of the steel pipe during processing.

[0109] S122. Convert the outer diameter of the finished pipe at normal temperature into the outer diameter of hot red steel by adding the thermal expansion compensation amount. The specific calculation method is shown in Table 2:

[0110] Table 2

[0111]

[0112] S123. Perform Z-score standardization processing on the data in the initial sample.

[0113] S2. Based on the preprocessed data, use a random forest model to calculate the importance of each preliminary feature, select important features according to the importance, and generate training data based on the important features.

[0114] Optionally, the steps of using a random forest model in S2 to calculate the importance of each preliminary feature, select important features according to the importance, and generate training data based on the important features may include the following steps S21 - S25:

[0115] S21. Use the original feature data to train a random forest model to obtain the baseline performance, calculate the initial feature importance of each feature, and the parameters used by the random forest model are obtained by grid search. The set parameter value ranges and optimal values are shown in Table 3:

[0116] Table 3

[0117]

[0118] S22. Select a certain feature and randomly rearrange all the values of this feature, that is, shuffle or permute the feature values. The traditional random forest feature screening method usually uses the Gini coefficient evaluation method to calculate the feature importance. This method only considers the influence of a single feature on the outer diameter. However, in the seamless steel pipe hot rolling production process, there are problems of multi - variables, strong coupling, and non - linearity. There may be interaction effects between various features, jointly affecting the outer diameter of the steel pipe. Therefore, only using the Gini coefficient to calculate the importance of each feature is not sufficient to solve this problem. Introduce the concept of permutation importance to improve the Gini coefficient. By randomly scrambling the order of features and destroying the relationship between features and the outer diameter, the importance of each feature can be more accurately evaluated.

[0119] S23. Use the shuffled data set to test on the random forest model and recalculate the importance of this feature.

[0120] S24. Calculate the importance score of the feature by comparing the difference between the performance of the model after permuting the feature and the baseline performance. The greater the difference, the higher the importance of the feature.

[0121] Optionally, the calculation of the importance score of the preliminary feature in S24 is shown in the following formula (1):

[0122] (1)

[0123] In the formula, represents the importance score of the preliminary feature, represents the importance of the preliminary feature, represents the importance of the preliminary feature after rearrangement.

[0124] S25. Repeat steps S22 to S24 to evaluate each feature to obtain the importance scores of all features, and eliminate the features with lower importance. The importance scores of each feature are as Figure 3 shown, and the features in the final training samples are shown in Table 4:

[0125] Table 4

[0126]

[0127] S3. Build a data-driven seamless steel pipe outer diameter prediction model using an SVR model based on a hybrid kernel function. Train the seamless steel pipe outer diameter prediction model according to the training data to obtain a trained seamless steel pipe outer diameter prediction model.

[0128] Optionally, building a data-driven seamless steel pipe outer diameter prediction model using an SVR model based on a hybrid kernel function and training the seamless steel pipe outer diameter prediction model according to the training data in S3 may include the following steps S31 - S32:

[0129] S31. Build a data-driven seamless steel pipe outer diameter prediction model using an SVR (Support Vector Regression) model based on a hybrid kernel function. The kernel functions in the support vector regression machine (SVR) can be mainly divided into two types: local kernel functions and global kernel functions. Local kernel functions mainly focus on the relationship between adjacent sample data, emphasizing the local similarity between samples, which results in samples closer in distance contributing more to the kernel function, while samples at a greater distance have less impact on the kernel function. In contrast, global kernel functions mainly consider the global relationship between samples and take into account the contributions of all samples during the construction of the decision boundary. Among common kernel functions, the RBF kernel function is a typical local kernel function, while the Poly kernel function is a representative of global kernel functions. By combining multiple kernel functions in a linear combination, the newly constructed hybrid kernel function can balance the learning ability and generalization ability, thereby improving the prediction performance and applicability of the seamless steel pipe outer diameter prediction model as a whole. As Figure 4 shown, compared with the SVR model using a single kernel function and other representative models, the SVR model that mixes the RBF kernel function and the Poly kernel function is significantly superior to other models in terms of prediction accuracy and stability. Its deviation distribution is extremely concentrated, the median is almost zero, and there are very few outliers, indicating that the model has a consistent and accurate prediction ability for outer diameter values and can more effectively meet the high requirements for accuracy and stability in outer diameter prediction.

[0130] Among them, the hybrid kernel function is based on the Poly kernel function and the RBF kernel function and is constructed according to the following formula (2):

[0131] (2)

[0132] In the formula, represents the mixed kernel function, represents the mixing coefficient, which controls the weights of the RBF kernel function and the Poly kernel function in the mixed kernel function, represents the RBF kernel function expression, represents the Poly kernel function expression.

[0133] Among them, the Poly kernel function expression is:

[0134] (3)

[0135] In the formula: is the dimension of the Poly kernel function, that is, the polynomial degree.

[0136] The RBF kernel function expression is:

[0137] (4)

[0138] In the formula: is the bandwidth coefficient of the RBF kernel function, which controls the distribution of data in the high-dimensional space.

[0139] S32. Train the seamless steel pipe outer diameter prediction model according to the training data, and use the PSO algorithm to optimize the hyperparameters of the SVR model based on the mixed kernel function. The hyperparameters to be optimized include: regularization coefficient C, RBF kernel function coefficient σ, misclassification tolerance ε, and mixing coefficient λ.

[0140] The training samples, specific data are shown in Table 5:

[0141] Table 5

[0142]

[0143] The optimization range and optimization results are shown in Table 6:

[0144] Table 6

[0145]

[0146] Furthermore, input the test set data into the trained seamless steel pipe outer diameter prediction model to obtain the predicted values of the finished pipe outer diameter, and evaluate the model prediction results. Specifically, use the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R2) as the evaluation indicators of the model error to evaluate the prediction performance of the model.

[0147] Specifically, the mean absolute error (MAE) is calculated according to the following formula:

[0148] (5)

[0149] The root mean square error (RMSE) is calculated according to the following formula:

[0150] (6)

[0151] The coefficient of determination (R2) is calculated according to the following formula:

[0152] (7)

[0153] Among them, the above prediction results are shown in Table 7. The results show that the SVR model with a hybrid kernel function shows significant superiority in various performance indicators. Its RMSE, MAE, and R² are 0.1487, 0.1135, and 0.9842 respectively, which are significantly better than other models. In contrast, the RMSE, MAE, and R² of the SVR model with a Poly kernel function are 0.3313, 0.2108, and 0.9218 respectively, with relatively large errors, mainly because it pays too much attention to global relationships and ignores the learning of local similarities; the RMSE, MAE, and R² of the SVR model with an RBF kernel function are 0.2512, 0.1678, and 0.9550 respectively. Although the accuracy is improved compared with the Poly kernel function, due to the too strong local learning ability, the generalization ability is still insufficient. The SVR model with a hybrid kernel function balances the local learning ability and global generalization ability of the model by integrating the global learning ability of the Poly kernel function and the local fitting ability of the RBF kernel function, combined with precise hyperparameter tuning. In addition, compared with other prediction models, the SVR model with a hybrid kernel function also shows obvious advantages. The RMSE, MAE, and R² of the BP neural network are 0.2126, 0.1427, and 0.9678 respectively. Although it can capture non-linear relationships, its accuracy is still inferior to that of the SVR model with a hybrid kernel function; while the RMSE, MAE, and R² of the AdaBoost prediction model are 0.3178, 0.2541, and 0.9280 respectively, with relatively large errors and relatively low goodness of fit. Generally speaking, the SVR model with a hybrid kernel function not only far exceeds other models in terms of accuracy, but also shows particularly outstanding performance in terms of generalization ability and robustness, and can better adapt to the characteristics of complex data and noise interference in actual production.

[0154] Table 7

[0155]

[0156] S4. Obtain the data of the seamless steel pipe to be predicted, input it into the trained seamless steel pipe outer diameter prediction model, and obtain the prediction result of the seamless steel pipe outer diameter.

[0157] In the embodiments of the present invention, considering the comprehensive influence of the process parameters in the three major deformation stages of piercing, tube rolling, and sizing on the outer diameter of the steel pipe, preliminary screening of characteristic variables is carried out based on the mechanism knowledge of hot rolling of steel pipes and data analysis. By adding the downtime and historical rolling length, the influence of roll thermal deformation and roll wear during the rolling process is considered. Finally, using the idea of permutation importance through the random forest model, the importance degree of each characteristic variable is calculated, redundant variables are eliminated, and a dataset with high quality is obtained, which helps to improve the accuracy of the seamless steel pipe outer diameter prediction model.

[0158] The present invention optimizes the SVR model by using the method of mixing the Poly kernel function and the RBF kernel function, which is beneficial to improving the prediction accuracy and generalization ability of the model.

[0159] The present invention predicts the outer diameter of the finished pipe based on the process parameters of the three deformation processes of piercing, tube rolling, and sizing, which can be used for the diagnosis of the outer diameter deviation of hot-rolled seamless steel pipes and the optimization of process parameters, and is beneficial to eliminating the hysteresis of process parameter adjustment.

[0160] Figure 5 It is a block diagram of a seamless steel pipe outer diameter prediction device based on a hybrid kernel function SVR model shown according to an exemplary embodiment. This device is used for the seamless steel pipe outer diameter prediction method based on the hybrid kernel function SVR model. Refer to Figure 5 , this device includes an acquisition model 310, a generation model 320, a training model 330, and an output model 340. Among them:

[0161] The acquisition model 310 is used to select the preliminary features that have an impact on the outer diameter of the seamless steel pipe, acquire the relevant data of the preliminary features, and preprocess the relevant data to obtain the preprocessed data.

[0162] The generation model 320 is used to calculate the importance degree of each preliminary feature according to the preprocessed data by using the random forest model, select the important features according to the importance degree, and generate the training data according to the important features.

[0163] The training model 330 is used to construct a data-driven seamless steel pipe outer diameter prediction model by using the SVR model based on the hybrid kernel function, and train the seamless steel pipe outer diameter prediction model according to the training data to obtain a trained seamless steel pipe outer diameter prediction model.

[0164] The output model 340 is used to acquire the seamless steel pipe data to be predicted, input it into the trained seamless steel pipe outer diameter prediction model, and obtain the seamless steel pipe outer diameter prediction result.

[0165] In the embodiments of the present invention, considering the comprehensive influence of the process parameters in the three major deformation stages of piercing, tube rolling, and sizing on the outer diameter of the steel pipe, preliminary screening of characteristic variables is carried out based on the mechanism knowledge of hot rolling of steel pipes and data analysis. By adding the shutdown time and historical rolling length, the influence of roll thermal deformation and roll wear during the rolling process is considered. Finally, using the idea of permutation importance through the random forest model, the importance degree of each characteristic variable is calculated, redundant variables are eliminated, and a dataset with high quality is obtained, which helps to improve the accuracy of the seamless steel pipe outer diameter prediction model.

[0166] The present invention optimizes the SVR model by using a method combining the hybrid Poly kernel function and the RBF kernel function, which is beneficial to improving the prediction accuracy and generalization ability of the model.

[0167] The present invention predicts the outer diameter of the finished pipe according to the process parameters in the three deformation processes of piercing, tube rolling, and sizing, which can be used for the diagnosis of the outer diameter deviation of hot-rolled seamless steel pipes and the optimization of process parameters, and is beneficial to eliminating the hysteresis of process parameter adjustment.

[0168] Figure 6 is a schematic structural diagram of a seamless steel pipe outer diameter prediction device provided by an embodiment of the present invention, as Figure 6 shown, the seamless steel pipe outer diameter prediction device may include the above-mentioned Figure 5 shown seamless steel pipe outer diameter prediction device based on the hybrid kernel function SVR model. Optionally, the seamless steel pipe outer diameter prediction device 410 may include a first processor 2001.

[0169] Optionally, the seamless steel pipe outer diameter prediction device 410 may further include a memory 2002 and a transceiver 2003.

[0170] Among them, the first processor 2001 is connected to the memory 2002 and the transceiver 2003, such as through a communication bus.

[0171] Next, in combination with Figure 6 each component of the seamless steel pipe outer diameter prediction device 410 will be specifically introduced:

[0172] Among them, the first processor 2001 is the control center of the seamless steel pipe outer diameter prediction device 410, which can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPUs), or can be an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention, such as: one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs).

[0173] Optionally, the first processor 2001 can execute various functions of the seamless steel pipe outer diameter prediction device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0174] In a specific implementation, as an embodiment, the first processor 2001 can include one or more CPUs, such as Figure 6 the CPU0 and CPU1 shown in

[0175] In a specific implementation, as an embodiment, the seamless steel pipe outer diameter prediction device 410 can also include multiple processors, such as Figure 6 the first processor 2001 and the second processor 2004 shown in

[0176] Among them, the memory 2002 is used to store software programs for implementing the solution of the present invention and is controlled by the first processor 2001 for execution. The specific implementation method can refer to the above method embodiments and will not be elaborated here.

[0177] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, or may also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, or any other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently and be coupled to the first processor 2001 through an interface circuit ( Figure 6 not shown) of the seamless steel pipe outer diameter prediction device 410. The embodiments of the present invention do not make specific limitations in this regard.

[0178] The transceiver 2003 is used to communicate with a network device or with a terminal device.

[0179] Optionally, the transceiver 2003 may include a receiver and a transmitter ( Figure 6 not separately shown). Among them, the receiver is used to implement the receiving function, and the transmitter is used to implement the sending function.

[0180] Optionally, the transceiver 2003 may be integrated with the first processor 2001 or may exist independently and be coupled to the first processor 2001 through an interface circuit ( Figure 6 not shown) of the seamless steel pipe outer diameter prediction device 410. The embodiments of the present invention do not make specific limitations in this regard.

[0181] It should be noted that Figure 6 the structure of the seamless steel pipe outer diameter prediction device 410 shown does not constitute a limitation on the router. The actual knowledge structure recognition device may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0182] In addition, the technical effects of the seamless steel pipe outer diameter prediction device 410 may refer to the technical effects of the seamless steel pipe outer diameter prediction method based on the hybrid kernel function SVR model described in the above method embodiments, and will not be elaborated here.

[0183] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0184] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0185] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, or a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0186] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood with reference to the context.

[0187] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0188] It should be understood that in various embodiments of the present invention, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0189] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in connection with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present invention.

[0190] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0191] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0192] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0193] In addition, the functional units in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0194] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0195] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for predicting the outer diameter of a seamless steel pipe based on a hybrid kernel function SVR model, characterized in that: The method comprises: S1. Selecting preliminary features that affect the outer diameter of the seamless steel pipe, obtaining relevant data of the preliminary features, and preprocessing the relevant data to obtain preprocessed data; S2. Calculate the importance of each preliminary feature using a random forest model based on the preprocessed data, select important features based on the importance, and generate training data based on the important features; S3, using the SVR model based on the hybrid kernel function to build a data-driven seamless steel pipe outer diameter prediction model, and training the seamless steel pipe outer diameter prediction model according to the training data to obtain a trained seamless steel pipe outer diameter prediction model; S4. Obtain the seamless steel pipe data to be predicted, input it into the trained seamless steel pipe outer diameter prediction model, and obtain the seamless steel pipe outer diameter prediction result.

2. The method for predicting the outer diameter of a seamless steel pipe based on a hybrid kernel function SVR model according to claim 1, characterized in that: The selection of S1 has an impact on the outer diameter of the seamless steel pipe, including: Based on the knowledge of the mechanisms of the piercing stage, the tube rolling stage and the diameter reduction stage, the process parameters that affect the outer diameter of the seamless steel tube are selected; taking the process parameters, downtime and historical rolling length as preliminary features; The process parameters include the furnace discharge temperature, roll gap, plug diameter, plug length, top extension, feed angle, rolling angle, guide plate distance, roll speed, and roll bite speed in the piercing stage; the throat diameter, mandrel diameter, mandrel speed, feed angle, rolling angle, and roll speed in the tube rolling stage; the last stand hole type, number of stands, roll speed of each stand, and incoming steel type and specification data in the diameter reducing stage.

3. The method for predicting the outer diameter of a seamless steel pipe based on a hybrid kernel function SVR model according to claim 2, characterized in that: The downtime and historical rolling length are used to consider the influence of roller thermal deformation and roller wear on the outer diameter of the steel pipe during the rolling process; The process of obtaining the downtime includes: Based on the time point when the head of the steel pipe enters the preset position, the production interval between each steel pipe and the previous steel pipe is calculated to obtain the downtime of each steel pipe; The process of obtaining the historical rolling length includes: The historical rolling length of the first steel pipe produced in the calibration collection data is 0, and the historical rolling length of each steel pipe thereafter is equal to the historical rolling length of the previous steel pipe plus the tube billet length of the previous steel pipe, thereby obtaining the historical rolling length of each steel pipe.

4. The method for predicting the outer diameter of a seamless steel pipe based on a hybrid kernel function SVR model according to claim 1, characterized in that: The step of obtaining the relevant data of the preliminary features and preprocessing the relevant data to obtain preprocessed data in S1 includes: S11. According to the time recorded by the tracking system, the process parameter curve of each steel pipe is intercepted; the average or maximum value of the process parameter curve is taken to generate production process parameter data; the tooling data, basic information and the production process parameter data are matched according to the steel pipe number, and the relevant data of the preliminary characteristics is generated according to the time recorded by the diameter gauge; S12, performing data transformation processing and standardization processing on the relevant data to obtain pre-processed data.

5. The method for predicting the outer diameter of a seamless steel pipe based on a hybrid kernel function SVR model according to claim 1, characterized in that: The step S2 of using a random forest model to calculate the importance of each preliminary feature, selecting important features according to the importance, and generating training data according to the important features includes: S21, use the preprocessed data to train the random forest model, and calculate the importance of each preliminary feature according to the trained forest model; S22, selecting a preliminary feature, and randomly rearranging all feature values ​​of the selected preliminary feature; S23, calculating the importance of the rearranged preliminary features according to the trained forest model; S24, calculating the importance score of the preliminary feature by comparing the difference between the importance of the preliminary feature and the importance of the rearranged preliminary feature; S25, determine whether each preliminary feature is selected, if so, generate training data according to the importance score of the preliminary feature; if not, go to step S22.

6. The method for predicting the outer diameter of a seamless steel pipe based on a hybrid kernel function SVR model according to claim 5, characterized in that: The importance score of the preliminary feature is calculated in S24 as shown in the following formula (1): （1） In the formula, represents the importance score of the preliminary features, Indicates the importance of preliminary features, Indicates the importance of the preliminary features after rearrangement.

7. The method for predicting the outer diameter of a seamless steel pipe based on a hybrid kernel function SVR model according to claim 1, characterized in that: The S3 uses the SVR model based on the hybrid kernel function to build a data-driven seamless steel pipe outer diameter prediction model, and trains the seamless steel pipe outer diameter prediction model according to the training data, including: S31. A data-driven seamless steel pipe outer diameter prediction model is constructed using the SVR model based on a hybrid kernel function. The hybrid kernel function is based on the Poly kernel function and the RBF kernel function and is constructed according to the following formula (2): （2） In the formula, represents the mixed kernel function, represents the mixing coefficient, represents the RBF kernel function expression, Represents the Poly kernel function expression; S32, training the seamless steel pipe outer diameter prediction model according to the training data, and optimizing the hyperparameters of the SVR model based on the hybrid kernel function by using a PSO algorithm; The hyperparameters include: regularization coefficient, RBF kernel function coefficient, misclassification tolerance and mixing coefficient.

8. A seamless steel pipe outer diameter prediction device based on a hybrid kernel function SVR model, the seamless steel pipe outer diameter prediction device based on a hybrid kernel function SVR model is used to implement the seamless steel pipe outer diameter prediction method based on a hybrid kernel function SVR model as claimed in any one of claims 1 to 7, characterized in that: The device comprises: Acquiring a model for selecting preliminary features that have an impact on the outer diameter of the seamless steel pipe, acquiring relevant data of the preliminary features, and preprocessing the relevant data to obtain preprocessed data; Generate a model for calculating the importance of each preliminary feature using a random forest model based on the preprocessed data, select important features based on the importance, and generate training data based on the important features; A training model is used to construct a data-driven seamless steel pipe outer diameter prediction model using an SVR model based on a hybrid kernel function, and to train the seamless steel pipe outer diameter prediction model according to the training data to obtain a trained seamless steel pipe outer diameter prediction model; The output model is used to obtain the seamless steel pipe data to be predicted, and input it into the trained seamless steel pipe outer diameter prediction model to obtain the seamless steel pipe outer diameter prediction result.

9. A seamless steel pipe outer diameter prediction device, characterized in that: The seamless steel pipe outer diameter prediction device comprises: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program codes, which can be called by a processor to execute the method according to any one of claims 1 to 7.

Images