Wire and rod mechanical properties prediction method and computer readable storage medium

Abstract

The present invention provides a method for predicting the mechanical properties of wire rods and a computer-readable storage medium, which belongs to the technical field of wire rod production. The method comprises the following steps: obtaining sample data, wherein the sample data comprises mechanical property prediction parameters of the wire rods and corresponding mechanical property test results; preprocessing the sample data, wherein the preprocessing comprises a plurality of preprocessing methods, and corresponding training data sets are obtained according to various preprocessing methods and combinations of various preprocessing methods; establishing a performance prediction model according to the sample data; using each training data set to train and evaluate the performance prediction model respectively, and obtaining a final model; inputting the prediction parameters of the predicted object into the final model, and obtaining the mechanical property prediction results of the predicted object. The present invention can preprocess the sample data, and evaluate it according to the preprocessing method, thereby improving the accuracy of the prediction results and thus improving the inspection efficiency. The production parameters can also be optimized with mechanical properties as the goal.

Description

Technical Field

[0001] The present invention relates to the technical field of wire rod production, and in particular to a wire rod mechanical property prediction method and a computer-readable storage medium. Background Art

[0002] Accurately predicting mechanical properties based on the chemical composition of steel, production process parameters, etc. is an important research topic in the steel industry. Previous research has mainly focused on the field of hot-rolled plates and strips. At present, a steel structure performance prediction and control model is established based on industrial big data, realizing the online prediction of hot-rolled plate and strip mechanical properties. However, in the field of online bars, due to the high temperature, dynamic and instantaneous characteristics of its production process, its performance prediction is more difficult.

[0003] In online bar production, actual sampling inspection is adopted. After the production is completed, it cannot be delivered directly to the user. It is necessary to wait for the inspection results to come out before actual delivery. In the absence of any prior knowledge of the product, each batch of products needs to be sampled on the same scale. If the factory qualification rate is to be improved, the sampling scale needs to be increased. This practice will greatly reduce the detection efficiency. Summary of the invention

[0004] In view of the above-mentioned shortcomings of the prior art, an object of the present invention is to provide a wire rod mechanical property prediction method and a computer-readable storage medium, so as to solve the problems of low wire rod detection efficiency in the prior art.

[0005] To achieve the above-mentioned and other related purposes, the present invention provides a method for predicting the mechanical properties of a wire rod, comprising the following steps:

[0006] Acquiring sample data, wherein the sample data includes mechanical property prediction parameters of the wire rod and corresponding mechanical property test results;

[0007] Preprocessing the sample data, wherein the preprocessing includes a basic preprocessing method and an optional preprocessing method, wherein the basic preprocessing method includes missing value processing and abnormal sample filtering, and the optional preprocessing method includes dimensionless conversion, sample balancing, feature selection, and feature dimension reduction;

[0008] Establishing a performance prediction model based on the sample data;

[0009] The sample data is processed and evaluated in turn by each basic preprocessing method to obtain the sample data processed by each basic preprocessing method,

[0010] The sample data processed by each basic preprocessing method is processed and evaluated by each optional preprocessing method in turn to obtain sample data processed by each optional preprocessing method;

[0011] Using the sample data processed by each of the optional preprocessing methods to train the performance prediction model to obtain a final model;

[0012] Inputting the prediction parameters of the prediction object into the final model to obtain the prediction results of the mechanical properties of the prediction object;

[0013] Among them, the missing value processing includes deletion method, mean filling method and random interpolation filling method, abnormal sample filtering includes process rule threshold filtering method, dimensionless includes standardization and normalization, sample balancing adopts oversampling, feature selection includes variance selection method and correlation coefficient method, feature dimension reduction includes discriminant analysis method and principal component analysis method;

[0014] When processing missing values ​​for the sample data, the sample data are processed respectively by using a deletion method, a mean filling method, and a random interpolation filling method to obtain three groups of first training data subsets, and the performance prediction models are trained and evaluated respectively by using the three groups of first training data subsets, and the first training data subset corresponding to the performance prediction model with the highest accuracy is selected as the first training data set;

[0015] Filtering abnormal samples from the first training data set to obtain a second training data set;

[0016] When the second training data set is dimensionlessly processed, standardization and normalization are respectively used to process the second training data set to obtain two sets of second training data subsets, and the performance prediction models are respectively trained and evaluated by the two sets of second training data subsets, and the second training data subset corresponding to the performance prediction model with the highest accuracy is selected as the third training data set;

[0017] Performing sample balancing processing on the third training data set to obtain a fourth training data set;

[0018] When performing feature selection on the fourth training data set, the fourth training data set is processed using a variance selection method and a correlation coefficient method respectively to obtain two sets of fourth training data subsets, and the fourth training data subset corresponding to the performance prediction model with the highest accuracy is selected as the fifth training data set;

[0019] When performing feature dimensionality reduction processing on the fifth training data set, feature dimensionality reduction processing is performed on the fifth training data set using a discriminant analysis method, a principal component analysis method, and an improved principal component analysis method to obtain three groups of fifth training data subsets, and the fifth training data subset corresponding to the performance prediction model with the highest accuracy is selected as the sixth training data set.

[0020] Optionally, the dimensions of the performance prediction parameters include steel grade specifications, chemical composition, heating furnace parameters and rolling process parameters.

[0021] Optionally, the model hit rate P is used for accuracy calculation;

[0022] The model hit rate P is the determination coefficient,

[0023]

[0024] Among them, Y_actual is the performance test value, Y_predict is the performance prediction value, and Y_mean is the mean of the performance test value;

[0025] Alternatively, the model hit rate P is: number of correct predictions / total number of predictions,

[0026] Among them, the performance prediction value and the performance test value are judged as qualified / unqualified respectively.

[0027] If the pass / fail judgment results of the performance prediction value and the performance test value are the same, then it is counted in the correct prediction number, otherwise it is counted in the wrong prediction number; or, if the difference between the performance prediction value and the performance test value is less than the preset difference, then it is counted in the correct prediction number, otherwise it is counted in the wrong prediction number.

[0028] Optionally, the performance test results include yield strength, tensile strength, elongation, elongation after fracture, strength-to-yield ratio, total elongation at maximum force, and super-yield ratio.

[0029] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, the method for predicting the mechanical properties of wire rods as described in any one of the above items is implemented.

[0030] As described above, the wire rod material mechanical property prediction method and computer-readable storage medium of the present invention have the following beneficial effects: the sample data can be preprocessed, evaluated according to the preprocessing method, and the accuracy of the prediction results can be improved, thereby improving the inspection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Shown is a flowchart of a method for predicting mechanical properties of wire rods according to an embodiment of the present invention.

[0032] Figure 2 Shown is a flowchart of sample data preprocessing in an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The following describes the embodiments of the present invention through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and 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 invention.

[0034] It should be noted that the diagram provided in the present embodiment only illustrates the basic concept of the present invention in a schematic manner, so the diagram only shows the components related to the present invention rather than drawing according to the number, shape and size of the components during actual implementation. The type, quantity and ratio of each component during actual implementation can be a random change, and the component layout type may also be more complicated. The structure, ratio, size, etc. illustrated in the drawings of the present specification are only used to match the content disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the limiting conditions that the present invention can implement, so they have no technical substantive significance. Any modification of the structure, change of the proportional relationship or adjustment of the size should still fall within the scope of the technical content disclosed by the present invention without affecting the effect that the present invention can produce and the purpose that can be achieved. At the same time, the terms such as "upper", "lower", "left", "right", "middle" and "one" quoted in this specification are only for the convenience of narration, and are not used to limit the scope of the present invention. The change or adjustment of its relative relationship should also be regarded as the scope of the present invention without substantially changing the technical content.

[0035] like Figure 1 As shown, this embodiment provides a method for predicting the mechanical properties of wire rods, comprising the following steps:

[0036] S1 obtains sample data, which includes mechanical property prediction parameters of wire rods and corresponding mechanical property test results.

[0037] The performance test results include yield strength, tensile strength, elongation, elongation after fracture, strength-to-yield ratio, total elongation at maximum force, and super-yield ratio.

[0038] S2 preprocesses the sample data, wherein the preprocessing includes a plurality of preprocessing methods, and corresponding training data sets are obtained according to the various preprocessing methods and the combination of the various preprocessing methods.

[0039] S3 establishes a performance prediction model based on the sample data, and uses various training data sets to train and evaluate the performance prediction model respectively to obtain a final model.

[0040] S4 inputs the prediction parameters of the prediction object into the final model to obtain the prediction results of the mechanical properties of the prediction object.

[0041] Step S2 includes the following sub-steps:

[0042] First, multiple preprocessing categories can be preset.

[0043] Then, the preprocessing methods for implementing the preprocessing category under one of the preprocessing categories are obtained.

[0044] Finally, the sample data is processed by each preprocessing method to obtain a training data subset under the preprocessing category. The training data subsets under all preprocessing categories are obtained in sequence to form various training data sets.

[0045] Specifically, all preprocessing categories are classified into basic preprocessing methods that are necessary for sample data and optional preprocessing methods that are selected according to data conditions. The basic preprocessing methods include missing value processing and abnormal sample filtering, and the optional preprocessing methods include dimensionless conversion, sample balancing, feature selection, and feature dimensionality reduction.

[0046] The missing value processing includes deletion, mean filling and random interpolation filling, abnormal sample filtering includes process rule threshold filtering, dimensionless includes standardization and normalization, sample balancing uses oversampling, feature selection includes variance selection and correlation coefficient, feature dimension reduction includes discriminant analysis and principal component analysis. The principal component analysis also includes improved principal component analysis.

[0047] like Figure 2 As shown, in this embodiment, the basic preprocessing method includes missing value processing and abnormal sample filtering, and the optional preprocessing methods include dimensionless conversion, sample balancing, feature selection, and feature dimensionality reduction.

[0048] The following example illustrates the process of preprocessing sample data:

[0049] (1) The sample data is the original data 00. The missing values ​​of the original data 00 are processed by deletion, mean filling and random interpolation filling respectively. After obtaining three sets of training data subsets 1-1, 1-2 and 1-3, the performance prediction models are trained and evaluated respectively, and the training data set 01 corresponding to the performance prediction model with the highest accuracy is selected.

[0050] (2) Filter abnormal samples based on training data set 01 to obtain training data set 02;

[0051] (3) Standardization and normalization are used to perform dimensionless processing on training data set 02, and two training data sets 3-1 and 3-2 are obtained. The performance prediction model is trained and evaluated using the above two training data subsets and training data set 02, and training data set 03 is obtained.

[0052] (4) Perform sample balancing on training data set 03 to obtain training data subset 4-1, and train and evaluate the performance prediction model together with training data set 03 to obtain training data set 04.

[0053] (5) Use the variance selection method and the correlation coefficient method to perform feature selection on the training data set 04 to obtain training data subsets 5-1 and 5-2, repeat the above model evaluation operation, and train and evaluate the performance prediction model together with the training data set 04 to obtain the training data set 05.

[0054] (6) Using the feature discriminant analysis method (LDA), principal component analysis method (PCA) and improved principal component analysis method (improved PCA), the training data set 05 is processed for feature dimensionality reduction to obtain training data subsets 6-1, 6-2 and 6-3. The performance prediction model is then trained and evaluated together with the training data set 05 to obtain the training data set 06.

[0055] The training data set 06 was used to train the mechanical property prediction model to obtain the final model.

[0056] The mechanical property prediction parameters may include multiple dimensions. In step S3, multiple mechanical property prediction models are established according to the dimensions in the mechanical property prediction parameters, and are trained and evaluated respectively to obtain the final model.

[0057] Specifically, the dimensions of the performance prediction parameters include steel grade specifications, chemical composition, heating furnace parameters and rolling process parameters. Chemical composition information includes the constituent elements and their proportions of wire rods such as Al, Als, As, B, C, Ca, Cr, Cu, Mn, Mo, N, Nb, Ni, P, S, Si, Ti, V, and Ceq.

[0058] The heating furnace parameters include the temperature and duration of each section, such as the furnace entry temperature, preheating section temperature, heating section temperature, soaking section temperature, furnace exit temperature, heating time, soaking time, furnace time, etc.

[0059] The rolling process parameters are divided into two major categories: bars and wires, which vary depending on the specific production line, including current, elongation, speed, temperature and other information.

[0060] For bars, the rolling process parameters mainly include sub-stand current, sub-stand elongation, sub-stand speed, flying shear temperature, pre-finishing rolling temperature, upper cooling bed temperature, water tank flow, and water tank temperature. For wire rods, the rolling process parameters mainly include sub-stand current, sub-stand elongation, sub-stand speed, stand temperature, finishing rolling inlet temperature, wire laying temperature, and coiling temperature.

[0061] In this embodiment, the training data set includes two dimensions. One is to establish different training data sets for different steel grades, and the other is to establish different training data sets by using different preprocessing and preprocessing combination methods.

[0062] In step S3, the trained performance prediction model is evaluated by using accuracy, and the trained performance prediction model with the highest accuracy is selected as the model. The accuracy can be calculated by using the model hit rate P.

[0063] Specifically, the model hit rate P may be a determination coefficient.

[0064] at this time,

[0065] Among them, Y_actual is the performance test value, Y_predict is the performance prediction value, and Y_mean is the mean of the performance test value.

[0066] Alternatively, the model hit rate P may be equal to the number of correct predictions / total number of predictions.

[0067] The performance prediction value and the performance test value are judged as qualified / unqualified respectively. If the qualified / unqualified judgment results of the performance prediction value and the performance test value are the same, it is a correct prediction, otherwise it is an incorrect prediction. Alternatively, if the difference between the performance prediction value and the performance test value is less than a preset difference, it is a correct prediction, otherwise it is an incorrect prediction.

[0068] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, the method for predicting the mechanical properties of wire rods as described in any one of the above items is implemented.

[0069] In summary, the wire rod mechanical property prediction method and computer-readable storage medium of the present invention have the following beneficial effects: the sample data can be preprocessed and evaluated according to the preprocessing method to improve the accuracy of the prediction results, thereby improving the inspection and testing efficiency; at the same time, the production parameters can be optimized with performance as the goal, so as to achieve stable quality production of rod and wire products.

[0070] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A method for predicting mechanical properties of wire rods, characterized in that: The following steps are involved: Acquiring sample data, wherein the sample data includes mechanical property prediction parameters of the wire rod and corresponding mechanical property test results; Preprocessing the sample data, wherein the preprocessing includes a basic preprocessing method and an optional preprocessing method, wherein the basic preprocessing method includes missing value processing and abnormal sample filtering, and the optional preprocessing method includes dimensionless conversion, sample balancing, feature selection, and feature dimension reduction; Establishing a performance prediction model based on the sample data; The sample data is processed and evaluated in turn by each basic preprocessing method to obtain the sample data processed by each basic preprocessing method, The sample data processed by each basic preprocessing method is processed and evaluated by each optional preprocessing method in turn to obtain sample data processed by each optional preprocessing method; Using the sample data processed by each of the optional preprocessing methods to train the performance prediction model to obtain a final model; Inputting the prediction parameters of the prediction object into the final model to obtain the prediction results of the mechanical properties of the prediction object; Among them, the missing value processing includes deletion method, mean filling method and random interpolation filling method, abnormal sample filtering includes process rule threshold filtering method, dimensionless includes standardization and normalization, sample balancing adopts oversampling, feature selection includes variance selection method and correlation coefficient method, feature dimension reduction includes discriminant analysis method and principal component analysis method; When processing missing values ​​for the sample data, the sample data are processed respectively by using a deletion method, a mean filling method, and a random interpolation filling method to obtain three groups of first training data subsets, and the performance prediction models are trained and evaluated respectively by using the three groups of first training data subsets, and the first training data subset corresponding to the performance prediction model with the highest accuracy is selected as the first training data set; Filtering abnormal samples from the first training data set to obtain a second training data set; When the second training data set is dimensionlessly processed, standardization and normalization are respectively used to process the second training data set to obtain two sets of second training data subsets, and the performance prediction models are respectively trained and evaluated by the two sets of second training data subsets, and the second training data subset corresponding to the performance prediction model with the highest accuracy is selected as the third training data set; Performing sample balancing processing on the third training data set to obtain a fourth training data set; When performing feature selection on the fourth training data set, the fourth training data set is processed using a variance selection method and a correlation coefficient method respectively to obtain two sets of fourth training data subsets, and the fourth training data subset corresponding to the performance prediction model with the highest accuracy is selected as the fifth training data set; When performing feature dimensionality reduction processing on the fifth training data set, feature dimensionality reduction processing is performed on the fifth training data set using a discriminant analysis method, a principal component analysis method, and an improved principal component analysis method to obtain three groups of fifth training data subsets, and the fifth training data subset corresponding to the performance prediction model with the highest accuracy is selected as the sixth training data set.

2. The wire rod mechanical property prediction method according to claim 1, characterized in that: The dimensions of the performance prediction parameters include steel grade specifications, chemical composition, heating furnace parameters and rolling process parameters.

3. The wire rod mechanical property prediction method according to claim 1, characterized in that The model hit rate P is used for accuracy calculation; The model hit rate P is the determination coefficient, Among them, Y_actual is the performance test value, Y_predict is the performance prediction value, and Y_mean is the mean of the performance test value; Alternatively, the model hit rate P is: number of correct predictions / total number of predictions, Among them, the performance prediction value and the performance test value are judged as qualified / unqualified respectively. If the pass / fail judgment results of the performance prediction value and the performance test value are the same, then it is counted in the correct prediction number, otherwise it is counted in the wrong prediction number; or, if the difference between the performance prediction value and the performance test value is less than the preset difference, then it is counted in the correct prediction number, otherwise it is counted in the wrong prediction number.

4. The wire rod mechanical property prediction method according to claim 1, characterized in that: The performance test results include yield strength, tensile strength, elongation, elongation after fracture, strength-to-yield ratio, total elongation at maximum force, and super-yield ratio.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for predicting the mechanical properties of wire rods as claimed in any one of claims 1 to 4 is implemented.

Images