Converter tapping curve presetting method and device based on data mining

Abstract

The invention discloses a converter tapping curve presetting method and device based on data mining. The converter tapping curve presetting method comprises the steps that converter tapping process data are collected; performing data preprocessing through data cleaning, data specification, data integration and data conversion methods, and establishing a sample set; randomly dividing a sample set into a training set, a verification set and a test set through random sampling; selecting a data mining method, establishing a data mining model, and training the data mining model by using a training set, so that the data mining model outputs a converter tapping curve according to the input converter tapping process data; verifying the data mining model by using a verification set; evaluating the data mining model using a test set; and outputting a converter tapping curve through the qualified data mining model, and issuing the converter tapping curve to a converter tilting mechanism to control converter tapping. According to the method, the converter tapping curve is intelligently preset in a data mining mode, tapping slag and molten steel rephosphorization caused by unreasonable tapping curve are reduced, and the automatic tapping effect of the converter is improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of iron and steel metallurgy and intelligent manufacturing, and more specifically, to a method and device for presetting the tapping curve of a converter based on data mining. Background Art

[0002] With the rapid advancement of intelligent manufacturing in the iron and steel industry, on the basis of automation and informatization, digital, networked, and intelligent technologies have become the core driving forces for improving quality, increasing efficiency, reducing staff, and achieving green development. In the field of converter tapping technology, the traditional tapping method adopts a manual operation mode, and the converter tilting tapping is controlled according to manual experience and visual observation, relying on manual experience, resulting in low tapping efficiency; the automatic tapping method is based on the tapping time mode, tapping weight mode, and data model mode, and a tapping curve is set before tapping, which has certain improvement compared with manual tapping.

[0003] However, there are many factors affecting converter tapping, and the influence degrees of various factors on the tapping curve are different, resulting in great limitations in the above methods. Summary of the Invention

[0004] Technical Problems to be Solved by the Invention

[0005] Aiming at the defects and deficiencies existing in the prior art, the present invention provides a method for presetting the tapping curve of a converter based on data mining. Through data collection, data preprocessing, data mining, and model evaluation, considering influencing factors such as furnace age, tapping hole, molten steel weight, slag volume, rephosphorization rate, tapping time, etc., it overcomes the influence of equipment and process fluctuations such as furnace lining erosion, tapping hole erosion, and molten steel weight on the tapping process, provides data support for setting the tapping curve, improves the tapping effect of the converter, and reduces the fluctuation of molten steel quality.

[0006] Technical Solution

[0007] To achieve the above object, the technical solution provided by the present invention is as follows:

[0008] A method for presetting the tapping curve of a converter based on data mining, comprising the following steps:

[0009] Step 1: Collect the tapping process data of the converter;

[0010] Step 2: Perform data preprocessing through data cleaning, data reduction, data integration, and data transformation methods to establish a sample set;

[0011] Step 3: Randomly divide the sample set into a training set, a validation set, and a test set through random sampling;

[0012] Step 4: Establish multiple data mining models using various data mining methods, and train the multiple data mining models using the training set, so that each data mining model outputs a converter tapping curve according to the input converter tapping process data;

[0013] Step 5: Verify each data mining model using the validation set, and select the data mining model with the smallest generalization error as the final model;

[0014] Step 6: Evaluate the final model using the test set;

[0015] Step 7: Output the converter tapping curve through the final model that passes the evaluation, and send it to the converter tilting mechanism to control the converter tapping.

[0016] Optionally, in Step 1 above, as Figure 1 shown, data collection can be carried out using any one of the Modbus, MQTT, OPC, TCP / IP communication protocols, and data collection is performed through the fieldbus, industrial Ethernet, 5G network, respectively, in communication with MES, PCS, PLC, and sensors; the collected data includes but is not limited to steel grade, furnace age, number of times the tapping hole is used, molten steel weight, steel slag weight, tilting angle θ, holding time t, where the tapping curve is a function of the tilting angle θ and the holding time t.

[0017] Optionally, in Step 2, data cleaning: includes missing value processing and outlier processing. Among them, for missing values, when the sample set is greater than the set quantity, the sample is excluded; when the sample set is not greater than the set quantity, the missing value is replaced with the average or median; for outliers, outlier detection is performed through the 3σ principle or box plot method, and the outliers are treated as missing values;

[0018] Data integration: Taking the converter heat as a unit, the data is organically integrated, with one converter heat as a sample and all samples as the sample set;

[0019] Data reduction: Through principal component analysis PCA, the main feature components of the data are extracted for dimensionality reduction;

[0020] Data transformation: The data is transformed into a form suitable for data mining through standardization and smoothing.

[0021] Optionally, in Step 3, the random sampling method is used to divide the sample set into a training set, a validation set, and a test set. Among them, the training set is 50% of the sample set and is used to train the model; the validation set is 25% of the sample set and is used to select the final model; the test set is 25% of the sample set and is used to evaluate the final model;

[0022] Optionally, in step 4, data mining is performed according to process data such as furnace age, tapping hole, molten steel weight, and steel slag quality related to converter tapping. Multiple data mining models are established using various data mining methods including statistical methods, classification rules, association rules, clustering analysis, neural networks, and genetic algorithms. Each model is trained using a training set to fit the parameters.

[0023] Optionally, in step 5, a validation set is used to verify the generalization ability of the data mining model and fit the hyperparameters.

[0024] Optionally, in step 6, a test set is used to evaluate the final model. When the reliability of the final model meets the tapping process requirements, the training is stopped for use. When the reliability does not meet the process requirements, iterative training is performed by optimizing data samples, adjusting model parameters, and changing data mining methods until the tapping process requirements are met.

[0025] Optionally, in step 7, the qualified final model is deployed and applied. Before converter tapping, according to the furnace age, tapping hole, molten steel weight, and steel slag weight data, a converter tapping curve is preset and sent to the converter tilting mechanism to control converter tapping.

[0026] Optionally, the method for presetting the converter tapping curve updates the final model by automatically establishing new samples and updating the sample set.

[0027] This application also provides a device for presetting a converter tapping curve based on data mining, including:

[0028] A data acquisition module for collecting converter tapping process data;

[0029] A data preprocessing module for preprocessing data through data cleaning, data reduction, data integration, and data transformation methods to establish a sample set;

[0030] A data mining module that randomly divides the sample set into a training set, a validation set, and a test set by random sampling. Multiple data mining models are established using at least one data mining method, and the multiple data mining models are trained using the training set so that each data mining model outputs a converter tapping curve according to the input converter tapping process data;

[0031] A model evaluation module for using the validation set to select the data mining model with the smallest generalization error as the final model and evaluating the final model using the test set;

[0032] A model application module for outputting a converter tapping curve through the qualified final model and sending it to the converter tilting mechanism to control converter tapping.

[0033] Beneficial effects

[0034] Adopting the technical solution provided by the present invention, compared with the prior art, it has the following beneficial effects:

[0035] The present invention can establish the association rules between various process parameters of converter tapping and the tapping curve through data mining, overcome the problems of the original technology relying on manual experience and limited process parameters involved, intelligently preset the optimal tapping curve according to relevant data before converter tapping, reduce the degree of manual intervention, improve the hysteresis of feedback adjustment means such as infrared slag detection and ladle overflow detection at the furnace mouth, reduce slagging during tapping and rephosphorization of molten steel, and optimize the converter automatic tapping technology. Brief Description of the Drawings

[0036] Figure 1 It is the process flow chart of the converter tapping curve preset method based on data mining in the embodiment of the present invention;

[0037] Figure 2 It is the data mining flow chart in the embodiment of the present invention. Detailed Embodiments

[0038] The technical solution of the present invention will be clearly and completely described below with reference to the drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0039] The present invention will be further described below with reference to the drawings and embodiments:

[0040] A method for presetting the converter tapping curve based on data mining in this embodiment. The nominal capacity of the converter of a certain iron and steel enterprise is 150t, the smelting cycle is about 30min, and the operation mode of using the tapping hole at the back of the furnace for tapping and the slagging at the furnace mouth in front of the furnace is adopted. The steps of automatic tapping are carried out through the converter tapping curve preset method of the present invention as follows:

[0041] Step 1: Collect the furnace number, steel grade, furnace age, and the number of times the tapping hole is used through the steelmaking MES system (Manufacturing Enterprise Production Process Execution System); collect the molten steel weight and slag weight through the secondary steelmaking system; collect the converter tilting angle and holding time through the converter tilting control PLC and sensors.

[0042] Step 2: Perform data preprocessing, including data cleaning, data reduction, data integration, and data conversion.

[0043] Through data cleaning, missing value handling and outlier handling are carried out: for missing values, when the sample set > 1000, the sample is removed; when the sample set ≤ 1000, the missing values are replaced with the mean or median; the 3σ principle is used to detect outliers, and for outliers, when the sample set > 1000, the sample is removed; when the sample set ≤ 1000, the missing values are replaced with the mean or median.

[0044] Through data reduction, principal component analysis (PCA) is adopted for dimensionality reduction, and the attributes irrelevant to the data mining model are deleted;

[0045] Through data integration, taking the converter heat as the unit, the collected distributed multi-source heterogeneous data are organically integrated, and the relevant data corresponding to one heat are used as one sample, and all samples are used as the sample set.

[0046] Through data transformation, the data are converted into a form suitable for data mining by means of standardization, smoothing, etc.

[0047] Step Three: Adopt the random sampling method, as Figure 2 shown, the sample set is divided into a training set, a validation set and a test set. Among them, the training set is 50% of the sample set and is used to train the model; the validation set is 25% of the sample set and is used to adjust and select the final model; the test set is 25% of the sample set and is used to evaluate the final model;

[0048] Step Four: According to the converter tapping process, the model inputs are set as: one or more of steel grade, furnace age, tapping hole, molten steel weight, slag weight, etc., and the model output is: the converter tapping curve; according to the sample situation, using one or more of statistical methods, classification rules, association rules, clustering analysis, decision trees, fuzzy sets, neural networks, genetic algorithms, etc., a data mining model is established and the training set is used to train the model;

[0049] Step Five: Use the validation set to verify the model, continuously adjust the model according to the verification results, select the best-performing model by minimizing the generalization error, and then use the training set and the validation set to train the model as the final model;

[0050] Step Six: Use the test set to evaluate the final model, evaluate the correctness of the data mining results and the reliability of the model. When the reliability of the final model meets the process requirements, stop training for use; when its reliability does not meet the process requirements, through optimizing the data samples, adjusting the model parameters, etc., iterative training is carried out until the process requirements are met;

[0051] Step Seven: Deploy and apply the final model that has passed the test. Before the converter taps, according to data such as furnace age, tapping hole, molten steel weight, slag weight, etc., the converter tapping curve is intelligently preset and sent to the converter tilting mechanism to control the converter tapping;

[0052] Step Eight: Establish a new sample and update the sample set, and repeat the above steps to achieve self-learning.

[0053] Through the preset method of the converter tapping curve based on data mining, compared with the prior art, the present invention reduces the manual intervention degree in the tapping process by more than 20%, the automatic tapping utilization rate reaches more than 95%, the slag volume in tapping is reduced by more than 10%, the rephosphorization rate of molten steel is reduced to less than 0.002%, improves the desulfurization effect, shortens the desulfurization time, and improves the intelligent level.

[0054] Certainly, the present invention may also have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and deformations according to the present invention, but these corresponding changes and deformations all fall within the protection scope of the claims of the present invention.

Claims

1. A preset method for the tapping curve of a converter based on data mining, characterized in that It includes the following steps: Step 1: Collect the tapping process data of the converter; Step 2: Perform data preprocessing through data cleaning, data reduction, data integration, and data transformation methods to establish a sample set; Step 3: Randomly divide the sample set into a training set, a validation set, and a test set by random sampling; Step 4: Establish multiple data mining models using at least one data mining method, and use the training set to train the multiple data mining models so that each data mining model outputs a tapping curve of the converter according to the input tapping process data of the converter; Step 5: Use the validation set to verify each data mining model, and select the data mining model with the smallest generalization error as the final model; Step 6: Use the test set to evaluate the final model; Step 7: Output the tapping curve of the converter through the final model that passes the evaluation, and send it to the converter tilting mechanism to control the tapping of the converter.

2. The preset method for the tapping curve of a converter based on data mining according to claim 1, wherein In the above Step 1, data collection is carried out by using any one of the Modbus, MQTT, OPC, and TCP / IP communication protocols, and communicating with the MES, PCS, PLC, and sensors respectively through the fieldbus, industrial Ethernet, and 5G network; the collected data includes but is not limited to steel grade, furnace age, number of times the tapping hole is used, molten steel weight, slag weight, tilting angle θ, and holding time t, where the tapping curve is a function of the tilting angle θ and the holding time t.

3. The preset method of the converter tapping curve based on data mining according to claim 1, characterized in that, In the above Step 2, data cleaning: includes missing value processing and outlier processing. Among them, for missing values, when the sample set is greater than the set number, the sample is removed; when the sample set is not greater than the set number, the missing value is replaced with the average or median; for outliers, outlier detection is carried out by the 3σ principle or the box plot method, and the outliers are treated as missing values; Data integration: Integrate the data organically in units of converter heats, with one converter heat as a sample, and all samples as the sample set; Data reduction: Extract the main feature components of the data through principal component analysis PCA for dimensionality reduction; Data transformation: Convert the data into a form suitable for data mining through standardization and smoothing.

4. A preset method for the converter tapping curve based on data mining according to claim 1, characterized in that, In the above Step 3, the random sampling method is used to divide the sample set into a training set, a validation set, and a test set. Among them, the training set is 50% of the sample set and is used to train the model; the validation set is 25% of the sample set and is used to select the final model; the test set is 25% of the sample set and is used to evaluate the final model.

5. A preset method for the converter tapping curve based on data mining according to claim 1, characterized in that, In the above Step 4, data mining is based on the process data of furnace age, tapping hole, molten steel weight, and slag quality related to converter tapping, and uses at least one of statistical methods, classification rules, association rules, clustering analysis, neural networks, and genetic algorithms to establish multiple data mining models, and uses the training set to train each model to fit the parameters.

6. The preset method of the converter tapping curve based on data mining according to claim 5, wherein In the above Step 5, use the validation set to verify the generalization ability of the data mining model and fit the hyperparameters.

7. A preset method for the tapping curve of a converter based on data mining according to claim 1, characterized in that In step six, use the test set to evaluate the final model. When the reliability of the final model meets the requirements of the tapping process, stop training for use. When the reliability does not meet the process requirements, iteratively train by optimizing data samples, adjusting model parameters, and replacing data mining methods until the tapping process requirements are met.

8. A preset method for the tapping curve of a converter based on data mining according to claim 1, characterized in that, In step seven, deploy and apply the qualified final model. Before tapping in the converter, preset the converter tapping curve according to the furnace age, tapping hole, molten steel weight, and slag weight data, and send it to the converter tilting mechanism to control the converter tapping.

9. A preset method for the tapping curve of a converter based on data mining according to claim 1, characterized in that The converter tapping curve preset method updates the final model by automatically creating new samples and updating the sample set.

10. A preset device for the tapping curve of a converter based on data mining, characterized in that, It includes: A data acquisition module for collecting converter tapping process data; A data preprocessing module for preprocessing data through data cleaning, data reduction, data integration, and data transformation methods to establish a sample set; A data mining module that randomly divides the sample set into a training set, a validation set, and a test set by random sampling, uses at least one data mining method to establish multiple data mining models, and uses the training set to train the multiple data mining models so that each data mining model outputs a converter tapping curve according to the input converter tapping process data; A model evaluation module for using the validation set to select the data mining model with the smallest generalization error as the final model and using the test set to evaluate the final model; A model application module for outputting the converter tapping curve through the qualified final model and sending it to the converter tilting mechanism to control the converter tapping.

Images