Converter production data processing method, device, equipment and medium

Abstract

The invention relates to the technical field of ferrous metallurgy and data processing, and discloses a converter production data processing method, device, equipment and medium, and the method comprises the following steps: obtaining data values of various converter production data of each converter in a plurality of converters; performing abnormal value detection on the data values through data treatment, a classification model based on a decision tree and a self-adaptive box graph to obtain selected data values of various converter production data; and according to the selected data value, determining a correlation coefficient among the various converter production data, and mapping the correlation coefficient to a color space to obtain a heat map. The relevance among various converter production data is mined, and powerful support is provided for subsequent data analysis, model training, process guidance and the like. A classification model based on a decision tree, a self-adaptive box diagram and a heat map technology are combined, the intelligent processing efficiency of converter production data is greatly improved, and an advanced and reliable solution is provided for improving the production efficiency and optimizing decisions.

Description

Technical Field

[0001] The present invention relates to the field of iron and steel metallurgy and data processing technology, and in particular to a method, device, equipment and medium for processing converter production data. Background Art

[0002] With the rise of industry and intelligent manufacturing, the production data of steel enterprises has increased dramatically. How to effectively manage and utilize this production data has become a huge challenge facing steel enterprises.

[0003] Since the chemical reactions in the steelmaking process are complex and difficult to fully grasp, historical production data is needed to provide practical guidance for current smelting. However, for a large amount of historical production data, how to select useful data from these historical production data full of a lot of noise is a technical problem that steel companies continue to solve.

[0004] The existing methods for processing steelmaking data mainly focus on single data analysis and lack comprehensive analysis of multivariate data. This leads to insufficient mining of complex relationships between data when processing large-scale and multi-dimensional converter production data, and inability to deeply analyze the interrelationships between various production data.

[0005] Therefore, there is an urgent need for a method for processing converter production data that can effectively mine the relationship between multiple converter production data. Summary of the invention

[0006] The present invention provides a method, device, equipment and medium for processing converter production data. By regularly acquiring a variety of converter production data and eliminating abnormal data by means of data governance, a classification model based on a decision tree, an adaptive box plot and the like, the correlation between the various converter production data can be mined, thereby providing guidance for subsequent optimization of process parameters and improvement of production efficiency.

[0007] In a first aspect, a method for processing converter production data is provided, the method comprising: obtaining data values ​​of a plurality of converter production data for each of a plurality of converters; For data values, a first sample set is obtained through data governance, a first abnormal sample in the first sample set is eliminated using a classification model based on a decision tree to obtain a second sample set, and an adaptive box plot is used to eliminate the second abnormal sample in the second sample set to obtain selected data values ​​of various converter production data; Based on the selected data values, the correlation coefficients between various converter production data are determined, and the correlation coefficients are mapped to the color space to obtain a heat map.

[0008] In a second aspect, a device for processing converter production data is provided, the device comprising: An acquisition unit, used for acquiring data values ​​of a plurality of converter production data of each converter among the plurality of converters; A processing unit is used to obtain a first sample set through data governance for data values, remove a first abnormal sample from the first sample set using a classification model based on a decision tree to obtain a second sample set, and remove a second abnormal sample from the second sample set using an adaptive box plot to obtain selected data values ​​of multiple converter production data; The generating unit is used for determining the correlation coefficients between the various converter production data according to the selected data values, and mapping the correlation coefficients to the color space to obtain a heat map.

[0009] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method for processing converter production data when executing the computer program.

[0010] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned method for processing converter production data are implemented.

[0011] In the scheme implemented by the above-mentioned processing method, device, equipment and medium for converter production data, data values ​​of multiple converter production data of each converter in multiple converters are obtained; for the data values, a first sample set is obtained through data governance, and a first abnormal sample in the first sample set is eliminated using a classification model based on a decision tree to obtain a second sample set, and an adaptive box plot is used to eliminate the second abnormal sample in the second sample set to obtain selected data values ​​of multiple converter production data; according to the selected data values, the correlation coefficient between the multiple converter production data is determined, and the correlation coefficient is mapped to the color space to obtain a heat map. The present invention can mine the correlation between multiple converter production data, and provide strong support for subsequent data analysis, model training, process guidance, etc. The combination of the decision tree-based classification model, the adaptive box plot and the heat map technology greatly improves the intelligent processing efficiency of converter production data, and provides an advanced and reliable solution for improving production efficiency and optimizing decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0013] Figure 1 It is a schematic flow chart of a method for processing converter production data in one embodiment of the present invention; Figure 2 1 is a box plot of a second sample subset of molten iron quality, scrap steel quality, oxygen consumption, mass fraction of C in molten steel, mass fraction of P in molten steel, and mass fraction of S in molten steel in one embodiment of the present invention, wherein (a) is a box plot of a second sample subset of molten iron quality, (b) is a box plot of a second sample subset of scrap steel quality, (c) is a box plot of a second sample subset of oxygen consumption, (d) is a box plot of a second sample subset of mass fraction of C in molten steel, (e) is a box plot of a second sample subset of mass fraction of P in molten steel, and (f) is a box plot of a second sample subset of mass fraction of S in molten steel; Figure 3 1 is a box plot of a second sample subset of the mass fraction of C in molten iron, the mass fraction of Si in molten iron, the mass fraction of Mn in molten iron, the mass fraction of P in molten iron, and the mass fraction of S in molten iron in one embodiment of the present invention, wherein (g) is a box plot of a second sample subset of the mass fraction of C in molten iron, (h) is a box plot of a second sample subset of the mass fraction of Si in molten iron, (i) is a box plot of a second sample subset of the mass fraction of Mn in molten iron, (j) is a box plot of a second sample subset of the mass fraction of P in molten iron, and (k) is a box plot of a second sample subset of the mass fraction of S in molten iron; Figure 4 is a heat map of various converter production data in one embodiment of the present invention; Figure 5 is a schematic diagram of the structure of a device for processing converter production data in one embodiment of the present invention; Figure 6 It is a schematic diagram of the structure of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0015] The idea of ​​the present invention is to introduce a method for processing converter production data based on a classification model, an adaptive box plot and a heat map in view of the deficiencies in the prior art in the intelligent processing and analysis of converter production data. The classification model based on the decision tree and the adaptive box plot can accurately identify and process outliers, and the heat map technology can explore the relationship between data, thereby providing guidance for subsequent optimization of process parameters and improvement of production efficiency.

[0016] In one embodiment, a method for processing converter production data is provided. Figure 1 As shown, Figure 1 A schematic flow chart of a method for processing converter production data provided by an embodiment of the present invention comprises the following steps: Step S110, obtaining data values ​​of various converter production data of each of the multiple converters.

[0017] The present invention first obtains data values ​​of a plurality of converter production data of each of a plurality of converters.

[0018] During the steelmaking process, a sensor may be provided on the converter to collect data values ​​of the converter production data of the converter through the sensor. The number of converters may be multiple, and each converter is identified by a unique furnace number. The converter production data of each converter may be multiple, for example, the converter production data may include but is not limited to the following types: molten iron quality, scrap steel quality, slag iron quality, oxygen consumption, mass fraction of C in molten steel, mass fraction of P in molten steel, mass fraction of S in molten steel, mass fraction of C in molten iron, mass fraction of Si in molten iron, mass fraction of Mn in molten iron, mass fraction of P in molten iron, mass fraction of S in molten iron, mass fraction of V in molten iron, mass fraction of Ti in molten iron, mass fraction of Pb in molten iron, mass fraction of Zn in molten iron, etc. The data values ​​of each converter production data are obtained.

[0019] Step S120, for the data values, a first sample set is obtained through data governance, a first abnormal sample in the first sample set is eliminated using a classification model based on a decision tree to obtain a second sample set, and an adaptive box plot is used to eliminate the second abnormal sample in the second sample set to obtain selected data values ​​of multiple converter production data.

[0020] The present invention then eliminates abnormal values ​​from the data values ​​of various converter production data.

[0021] Data governance is performed on data values. Data governance may include, but is not limited to: data type verification, unit consistency verification, data format verification, etc. Data type verification aims to ensure that data values ​​are numerical values, unit consistency verification aims to ensure that the units of data values ​​are unified, and data format verification aims to ensure that data values ​​are within a reasonable and valid numerical range. Through data governance, the first sample set for preliminary screening of abnormal values ​​is obtained.

[0022] The first abnormal sample is eliminated from the first sample set using a classification model based on a decision tree. The classification model based on a decision tree may be a pre-trained machine learning algorithm, which determines whether a data value in the first sample set is an abnormal value through cluster analysis. The abnormal value is further screened through the output of the classification model based on the decision tree, thereby obtaining a second sample set.

[0023] The second abnormal sample is removed from the second sample set using an adaptive box plot. The adaptive box plot is to introduce an adaptive variable into the definition of the outlier boundary of the box plot, and the adaptive variable can adjust the outlier boundary of the box plot according to the historical production data. Through the adaptive box plot, the outliers are screened again, thereby obtaining a variety of selected production data.

[0024] Step S130, determining the correlation coefficients between the various converter production data based on the selected data values, and mapping the correlation coefficients to the color space to obtain a heat map.

[0025] The present invention finally determines the correlation and generates a heat map.

[0026] Based on the selected data values ​​of various converter production data, the correlation between each converter production data can be calculated. And the correlation is converted into corresponding colors. The colors and correlations are plotted in a chart to obtain an intuitive heat map.

[0027] from Figure 1 It can be seen from the method shown that the method for processing converter production data provided by the present invention obtains data values ​​of multiple converter production data of each converter in multiple converters; for the data values, a first sample set is obtained through data governance, and a first abnormal sample in the first sample set is eliminated using a classification model based on a decision tree to obtain a second sample set, and an adaptive box plot is used to eliminate the second abnormal sample in the second sample set to obtain selected data values ​​of multiple converter production data; according to the selected data values, the correlation coefficient between the multiple converter production data is determined, and the correlation coefficient is mapped to the color space to obtain a heat map. The present invention can mine the correlation between multiple converter production data, and provide strong support for subsequent data analysis, model training, process guidance, etc. The combination of the decision tree-based classification model, the adaptive box plot and the heat map technology greatly improves the intelligent processing efficiency of converter production data, and provides an advanced and reliable solution for improving production efficiency and optimizing decision-making.

[0028] The present invention is described in detail below through specific embodiments.

[0029] First, data values ​​of a plurality of converter production data of each of a plurality of converters are acquired.

[0030] In some optional embodiments, step S110 obtains data values ​​of multiple converter production data of each of the multiple converters, including: using converter sensors to collect real-time values ​​of multiple converter production data of each of the multiple converters in real time; persistently storing the real-time values ​​in a database; and retrieving data values ​​to be analyzed from the database.

[0031] The real-time values ​​of various converter production data of each of the multiple converters can be collected in real time by using converter sensors installed at the multiple converters.

[0032] The real-time values ​​are persistently stored in the SQL database to ensure data reliability and long-term storage.

[0033] When analysis is required, the data values ​​that need to be analyzed are retrieved from the SQL database.

[0034] Table 1 Data values ​​of various converter production data for each of the multiple converters:

[0035] Table 1 shows the data values ​​of various converter production data of each converter in the multiple converters provided by the embodiment of the present invention. Table 1 is only an exemplary display of some furnace numbers, and the actual situation is not limited to this. In Table 1, except for the data values ​​of oxygen consumption, the unit is m 3 Except for , the unit of other data values ​​is ton (t), and N represents a non-number. In the case of the example in Table 1, multiple converters are distinguished by furnace number identification. Each converter includes: molten iron mass (expressed as "molten iron"), scrap steel mass (expressed as "scrap steel"), slag iron mass (expressed as "slag iron"), oxygen consumption, mass fraction of C in molten steel (expressed as "molten steel C"), mass fraction of P in molten steel (expressed as "molten steel P"), mass fraction of S in molten steel (expressed as "molten steel S"), mass fraction of C in molten iron (expressed as "molten iron C"), mass fraction of Si in molten iron (expressed as "molten iron Si"), mass fraction of Mn in molten iron (expressed as "molten iron Mn"), mass fraction of P in molten iron (expressed as "molten iron P"), mass fraction of S in molten iron (expressed as "molten iron S") converter production data.

[0036] Secondly, for the data values, the first sample set is obtained through data governance, and the first abnormal sample in the first sample set is eliminated using a decision tree-based classification model to obtain the second sample set. The second abnormal sample in the second sample set is eliminated using an adaptive box plot to obtain selected data values ​​of multiple production data.

[0037] In some optional embodiments, step S120 obtains a first sample set through data governance, including: eliminating non-integer and non-floating point data values; unifying the units of data values; eliminating data values ​​that exceed a normal threshold, wherein the normal threshold corresponds to the type of converter production data, and the normal threshold is determined based on the mean, standard deviation and control threshold sensitivity coefficient of the data values ​​of each converter production data; and the retained data values ​​are used as the first sample set.

[0038] The first sample set is obtained through data governance through the following steps: The first step is to remove non-integer and non-floating point data values.

[0039] Eliminating non-integer data values ​​can be achieved by using the following integer test formula (1): , formula (1).

[0040] Eliminating non-floating point data values ​​can be achieved by using the following floating point test formula (2): , formula (2).

[0041] That is, if a data value is neither an integer nor a floating point type, the data value will be discarded.

[0042] For example: In Table 1, for converter production data such as molten iron quality, data values ​​recorded as N will be eliminated; for converter production data such as scrap steel quality, data values ​​recorded as N will be eliminated; and the same applies to other types of converter production data.

[0043] The second step is to unify the units of data values.

[0044] The unit consistency of data values ​​can be achieved by the following unit consistency test formula (3): , formula (3).

[0045] That is, all data values ​​in kg are uniformly converted to data values ​​in t, and data values ​​other than the two (kg and t) remain unchanged.

[0046] For example: If there is a data value in kg, convert it to a data value in t.

[0047] The third step is to remove data values ​​that exceed the normal threshold.

[0048] Eliminating data values ​​that exceed the normal threshold can be achieved by using the following abnormal format test formula (4): , formula (4).

[0049] Among them, L represents the lower limit of the normal threshold, and U represents the upper limit of the normal threshold.

[0050] L is determined by the following normal threshold lower limit formula (5): , formula (5).

[0051] U is determined by the following normal upper threshold formula (6): , formula (6).

[0052] Among them, μ represents the mean value of the data value of a converter production data, σ represents the standard deviation of the data value of the converter production data, and m represents the control threshold sensitivity coefficient of the converter production data. The control threshold sensitivity coefficient is a numerical value flexibly determined artificially according to the accuracy requirements of data governance, which is a value greater than 0. When the accuracy requirements of data governance are high, the control threshold sensitivity coefficient can be smaller; when the accuracy requirements of data governance are low, the control threshold sensitivity coefficient can be larger. The control threshold sensitivity coefficient can be selected as a value between 1 and 1.5.

[0053] That is, if a data value is lower than the normal threshold lower limit or higher than the normal threshold upper limit of the corresponding type of converter production data, the data value will be eliminated.

[0054] For example: after eliminating non-integer and non-floating-point data and unifying the units, several data values ​​are retained for the converter production data of molten iron quality. Then, the mean μ_molten iron of the several data values ​​and the standard deviation σ_molten iron of the several data values ​​are calculated, and the control threshold sensitivity coefficient m_molten iron of the molten iron quality is determined, and then the normal threshold lower limit L_molten iron and the normal threshold upper limit U_molten iron of the molten iron quality are calculated, and the data values ​​lower than L_molten iron and higher than U_molten iron are eliminated; after eliminating non-integer data, the data values ​​lower than L_molten iron and higher than U_molten iron are eliminated. After the numeric type and non-floating point type are unified in unit, the converter production data of scrap steel quality retains several data values, then calculate the mean μ_scrap steel of the several data values, the standard deviation σ_scrap steel of the several data values, and determine the control threshold sensitivity coefficient m_scrap steel of scrap steel quality, and then calculate the normal threshold lower limit L_scrap steel and normal threshold upper limit U_scrap steel of scrap steel quality, and eliminate the data values ​​lower than L_scrap steel and higher than U_scrap steel from the several data values; the same applies to other types of converter production data.

[0055] The fourth step is to use the retained data values ​​as the first sample set.

[0056] After the data values ​​are processed by data governance, the retained data values ​​form a first sample set. The first sample set may include a plurality of first sample subsets corresponding to a plurality of converter production data.

[0057] For example, the data values ​​retained from converter production data such as molten iron quality constitute a first sample subset; the data values ​​retained from converter production data such as scrap steel quality constitute a first sample subset; and the same applies to other types of converter production data. Each first sample subset constitutes a first sample set.

[0058] After the above steps, the preliminary screening of converter production data is completed.

[0059] In some optional embodiments, step S120 uses a decision tree-based classification model to eliminate the first abnormal sample in the first sample set to obtain a second sample set, including: inputting the first sample set into the decision tree-based classification model; determining the first abnormal sample in the first sample set according to the output of the decision tree-based classification model; after eliminating the first abnormal sample, obtaining the second sample set; wherein the decision tree-based classification model is trained using historical data values ​​of multiple marked converter production data as training samples.

[0060] Obtaining the second sample set through the decision tree-based classification model is specifically achieved through the following steps: The first step is to train a classification model based on a decision tree.

[0061] The classification model based on the decision tree is trained by taking the historical data values ​​of various converter production data marked as training samples.

[0062] Historical values ​​of various converter production data for each of a plurality of converters are obtained from a SQL database.

[0063] Data governance is performed on historical values ​​to obtain historical data values. Data governance may include, but is not limited to, eliminating non-integer and non-floating point historical values, unifying the units of historical values, eliminating historical values ​​that exceed normal thresholds, etc.

[0064] Each historical data value is manually labeled as an abnormal value or a normal value.

[0065] According to the types of converter production data in the training samples and manual labeling as classification indicators, the classification tree is trained to obtain a classification model based on the decision tree.

[0066] In the second step, the first sample set is input into a classification model based on a decision tree, and a first abnormal sample in the first sample set is determined according to an output of the classification model based on the decision tree.

[0067] The first sample set is input into a classification model based on a decision tree, so that the classification model based on the decision tree outputs a prediction result of whether each data value in the first sample set is an abnormal value or a normal value. The data value output as the abnormal value prediction result is determined as the first abnormal sample.

[0068] The third step is to remove the first abnormal samples and obtain the second sample set.

[0069] After the first abnormal samples are removed, the remaining data values ​​form a second sample set. The second sample set may include a plurality of second sample subsets corresponding to a variety of converter production data.

[0070] After the above steps, the further screening of converter production data is completed.

[0071] In some optional embodiments, step S120 uses an adaptive box plot to eliminate second abnormal samples in the second sample set to obtain selected data values ​​of multiple converter production data, including: dividing the second sample set into multiple second sample subsets according to the type of converter production data; for each second sample subset, constructing a box according to the first quartile, median and third quartile of the second sample subset, determining the outlier boundary of the box according to the dynamic adjustment coefficient, and determining the data values ​​in the second sample subset that are outside the outlier boundary as second sub-abnormal samples; after eliminating multiple second sub-abnormal samples, selected data values ​​are obtained.

[0072] The use of adaptive box plots to obtain selected data values ​​for various converter production data is specifically achieved through the following steps: In the first step, the second sample set is divided into a plurality of second sample subsets according to the types of converter production data.

[0073] The second sample set is divided into a plurality of second sample subsets, each of which corresponds to each type of converter production data.

[0074] For example, the data values ​​of converter production data such as molten iron quality that are retained constitute a second sample subset; the data values ​​of converter production data such as scrap steel quality that are retained constitute a second sample subset; and the same applies to other types of converter production data. Each second sample subset constitutes a second sample set.

[0075] In the second step, for each second sample subset, boxes are constructed based on the first quartile, median, and third quartile of the second sample subset.

[0076] For each second sample subset, calculate the first quartile of the second sample subset , median and the third quartile , according to the first quartile , median and the third quartile Build the box.

[0077] For example: For the second sample subset corresponding to the molten iron quality, calculate its first quartile , median and the third quartile ,according to , and Construct a box corresponding to the molten iron quality; for the second sample subset corresponding to the scrap steel quality, calculate its first quartile , median and the third quartile ,according to , and Construct a box corresponding to the scrap steel mass; and so on for other types of converter production data.

[0078] The third step is to determine the outlier boundary of the box based on the dynamic adjustment coefficient.

[0079] The outlier boundaries include lower and upper boundaries.

[0080] The lower boundary is defined by the following formula (7): , formula (7).

[0081] The upper boundary is defined by the following formula (8): , formula (8).

[0082] Among them, K represents the dynamic adjustment coefficient, Indicates the length of the box. The dynamic adjustment coefficient K is a value determined according to the hard threshold index k and the soft threshold index α. The method for determining the dynamic adjustment coefficient K is described in detail below.

[0083] For example: For the box corresponding to the molten iron mass, its lower boundary is defined as , whose upper boundary is defined as ; For the box corresponding to the scrap mass, its lower boundary is defined as , whose upper boundary is defined as ; The production data of other types of converters are similar.

[0084] Figures 2 to 3 Box plots of multiple second sample subsets provided by embodiments of the present invention. Figures 2 to 3 As shown, the molten iron mass ( Figure 2 (a) in Figure 2), scrap steel quality ( Figure 2 (b) in Figure ), oxygen consumption ( Figure 2 (c) in Figure ), the mass fraction of C in the molten steel ( Figure 2 (d) in Figure ), mass fraction of P in molten steel ( Figure 2 (e) in Figure ), mass fraction of S in molten steel ( Figure 2 (f) in Figure ), mass fraction of C in molten iron ( Figure 3 (g) in Figure ), mass fraction of Si in molten iron ( Figure 3 (h) in Figure ), mass fraction of Mn in molten iron ( Figure 3 (i) in Figure 2), mass fraction of P in molten iron ( Figure 3 (j) in Figure ), mass fraction of S in molten iron ( Figure 3The box plot of the second sample subset in (k) of Figure 1 is shown in Figure 2. There is no box plot of slag quality because after the data values ​​of slag quality are processed by data governance and the classification model based on the decision tree to eliminate outliers, the remaining data values ​​are all 0, and there is no need to continue processing. Therefore, in this step and subsequent processing steps, the converter production data of slag quality is no longer processed.

[0085] In the fourth step, the data values ​​in the second sample subset that are outside the outlier boundary are determined as second sub-outlier samples.

[0086] Based on the abnormal value boundary defined by the box corresponding to each converter production data, the data values ​​in each second sample subset are judged as abnormal values. The data values ​​in each second sample subset that are outside the abnormal value boundary are determined as second sub-abnormal samples. Multiple second sub-abnormal samples form a second abnormal sample.

[0087] The fifth step is to remove multiple second sub-abnormal samples to obtain selected data values.

[0088] The second abnormal sample is removed from the second sample set, thereby obtaining selected data values ​​of a variety of converter production data.

[0089] After the above steps, the re-screening of the converter production data is completed.

[0090] For the third step above, the outlier boundary of the box is determined according to the dynamic adjustment coefficient. The dynamic adjustment coefficient includes: hard threshold index and soft threshold index; the hard threshold index is an empirical index; the soft threshold index is an adaptive index obtained by inputting the statistical index of the data value in the second sample subset into the regression model for prediction; wherein, the regression model corresponds to the type of converter production data, and the regression model is trained using the statistical index of the original data value of each type of marked converter production data as a training sample.

[0091] Two variables are introduced into the dynamic adjustment coefficient K. One variable is the hard threshold index k, and the other variable is the soft threshold index α. The relationship among the dynamic adjustment coefficient K, the hard threshold index k, and the soft threshold index α is expressed by the following formula (9): , formula (9).

[0092] The hard threshold index k is an empirical index. The value of the hard threshold index k can be manually determined by empirical values, or the hard threshold index k can be adjusted according to the feedback of the final heat map and other results. The hard threshold index k can be 1.5.

[0093] For example: for the second sample subset corresponding to molten iron quality, the hard threshold indicator k_molten iron is determined; for the second sample subset corresponding to scrap steel quality, the hard threshold indicator k_scrap steel is determined; and the same applies to other types of converter production data.

[0094] The soft threshold index α is an adaptive index obtained by inputting the statistical index of the data value in the second sample subset into the regression model for prediction. That is, the soft threshold index α is predicted by the regression model based on the statistical index of the data value in the second sample subset. The regression model is a pre-trained machine learning algorithm. The regression model corresponds to the type of converter production data and is trained using the statistical index of the original data value of each type of marked converter production data as a training sample.

[0095] The following describes the steps of determining the soft threshold indicator α_molten iron by taking the second sample subset corresponding to the molten iron quality as an example.

[0096] 1. Obtain the historical values ​​of molten iron quality of each of the multiple converters from the SQL database, eliminate outliers from the historical values ​​through data governance and a decision tree-based classification model to obtain the original data values.

[0097] 2. Calculate the statistical indicators of the original data values. The statistical indicators of the original data values ​​may include, but are not limited to: the mean of the original data values, the standard deviation of the original data values, the minimum value of the original data values, the maximum value of the original data values, the first quartile of the original data values, the median of the original data values, the third quartile of the original data values, the kurtosis of the original data values, and the skewness of the original data values.

[0098] 3. Manually mark the statistical indicators of the original data values.

[0099] 4. Using mean square error as the loss function and the statistical indicators of the labeled original data values ​​as training samples, the gradient descent algorithm is used to learn the parameters that minimize the loss function, thereby obtaining a regression model.

[0100] 5. Calculate the statistical indicators of the data values ​​of the second sample subset corresponding to the molten iron quality.

[0101] 6. The statistical index of the data value of the second sample subset is input into the regression model, and the regression model outputs the soft threshold index α_molten iron.

[0102] The same goes for other types of converter production data.

[0103] Table 2 shows the statistical indicators of the data values ​​of multiple second sample subsets provided by the embodiment of the present invention. Table 2 is only an exemplary display, and the actual situation is not limited to this. In Table 2, the statistical indicators of the second sample subset of molten iron quality (expressed as "molten iron"), scrap steel quality (expressed as "scrap steel"), slag iron quality (expressed as "slag iron"), oxygen consumption, mass fraction of C in molten steel (expressed as "molten steel C"), mass fraction of P in molten steel (expressed as "molten steel P"), mass fraction of S in molten steel (expressed as "molten steel S"), mass fraction of C in molten iron (expressed as "molten iron C"), mass fraction of Si in molten iron (expressed as "molten iron Si"), mass fraction of Mn in molten iron (expressed as "molten iron Mn"), mass fraction of P in molten iron (expressed as "molten iron P"), and mass fraction of S in molten iron (expressed as "molten iron S") are calculated respectively. In Table 2, the mean, standard deviation, minimum, maximum, first quartile (25%), median (50%), and third quartile (75%) are shown. Statistical indicators may also include kurtosis and skewness, which are not specifically shown in Table 2.

[0104] Table 2 Statistical indicators of data values ​​of multiple second sample subsets:

[0105] Finally, based on the selected data values, the correlation coefficients among various converter production data are determined, and the correlation coefficients are mapped to the color space to obtain a heat map.

[0106] In some optional embodiments, step S130 determines the correlation coefficients between multiple converter production data based on the selected data values, and maps the correlation coefficients to the color space to obtain a heat map, including: calculating the Pearson correlation coefficients between any two types of converter production data respectively; integrating the Pearson correlation coefficients into a correlation coefficient matrix; converting the correlation coefficient matrix into a color matrix; and forming a heat map based on the color matrix.

[0107] Through the previous steps, selected data values ​​of various converter production data are obtained.

[0108] For example: the quality of molten iron ultimately retains a selected data values, and the quality of scrap steel ultimately retains b selected data values; and so on for other types of converter production data.

[0109] In order to calculate the Pearson correlation coefficient between any two converter production data, the number of selected data values ​​of various converter production data must be equal. Therefore, data interpolation can be used to fill in the gaps.

[0110] The Pearson correlation coefficient between any two converter production data can be calculated using the following formula (10): , formula (10).

[0111] in, Indicates the first values, represents the mean of all values ​​of a converter production data after padding, Indicates the production data of another converter after completion. values, Represents the mean of all values ​​of another type of converter production data after padding.

[0112] The value range of is [-1,1]. That is, when When and Completely positive correlation; when When and Completely negative correlation; when When and Totally irrelevant.

[0113] After calculating the Pearson correlation coefficient of any two converter production data, it is integrated into a correlation coefficient matrix using the following formula (11): , formula (11).

[0114] in, Indicates the number of types of converter production data.

[0115] For example: Figures 2 to 3 In the case shown, The value of can be 11. The quality of molten iron can be defined as type number 1, and the quality of scrap steel can be defined as type number 2. Represents the Pearson correlation coefficient between molten iron quality and scrap steel quality; other types of converter production data are similar.

[0116] To convert the correlation matrix to a color matrix, define the color range . Indicates the lightest color. Indicates the darkest color.

[0117] The correlation coefficient matrix can be converted into a color matrix using the following formula (12): , formula (12).

[0118] When When , the corresponding color is the darkest ;when When the corresponding color is the lightest ;when , corresponding to neutral color.

[0119] Figure 4 A heat map of various converter production data provided by the embodiment of the present invention. Figure 4 As shown, Figure 4 A heat map showing the quality of molten iron (expressed as "molten iron"), the quality of scrap steel (expressed as "scrap steel"), oxygen consumption, the mass fraction of C in molten steel (expressed as "molten steel_C"), the mass fraction of P in molten steel (expressed as "molten steel_P"), the mass fraction of S in molten steel (expressed as "molten steel_S"), the mass fraction of C in molten iron (expressed as "molten iron_C"), the mass fraction of Si in molten iron (expressed as "molten iron_Si"), the mass fraction of Mn in molten iron (expressed as "molten iron_Mn"), the mass fraction of P in molten iron (expressed as "molten iron_P"), and the mass fraction of S in molten iron (expressed as "molten iron_S") is shown.

[0120] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0121] In one embodiment, a device for processing converter production data is provided. The device for processing converter production data corresponds to the method for processing converter production data in the above embodiment. Figure 5 As shown, the converter production data processing device 500 includes: An acquisition unit 510 is used to acquire data values ​​of a plurality of converter production data of each converter among a plurality of converters; The processing unit 520 is used to obtain a first sample set by data governance for the data value, remove a first abnormal sample from the first sample set by using a classification model based on a decision tree to obtain a second sample set, and remove a second abnormal sample from the second sample set by using an adaptive box plot to obtain selected data values ​​of multiple converter production data; The generating unit 530 is used to determine the correlation coefficients between the various converter production data according to the selected data values, and map the correlation coefficients to the color space to obtain a heat map.

[0122] In some optional embodiments, in the above-mentioned device 500, the acquisition unit 510 is specifically used to: use the converter sensor to collect the real-time values ​​of various converter production data of each converter in multiple converters in real time; store the real-time values ​​persistently in a database; and retrieve the data values ​​that need to be analyzed from the database.

[0123] In some optional embodiments, in the above-mentioned device 500, the processing unit 520 includes: a first processing module, used to: eliminate non-integer and non-floating point data values; unify the units of data values; eliminate data values ​​that exceed a normal threshold, wherein the normal threshold corresponds to the type of converter production data, and the normal threshold is determined based on the mean, standard deviation and control threshold sensitivity coefficient of the data values ​​of each converter production data; and use the retained data values ​​as the first sample set.

[0124] In some optional embodiments, in the above-mentioned device 500, the processing unit 520 includes: a second processing module, used to: input the first sample set into a classification model based on a decision tree; determine the first abnormal sample in the first sample set according to the output of the classification model based on the decision tree; after removing the first abnormal sample, obtain a second sample set; wherein the classification model based on the decision tree is trained using historical data values ​​of a variety of marked converter production data as training samples.

[0125] In some optional embodiments, in the above-mentioned device 500, the processing unit 520 includes: a third processing module, used to: divide the second sample set into multiple second sample subsets according to the type of converter production data; for each second sample subset, construct a box according to the first quartile, median and third quartile of the second sample subset, determine the outlier boundary of the box according to the dynamic adjustment coefficient, and determine the data values ​​in the second sample subset that are outside the outlier boundary as second sub-abnormal samples; after eliminating multiple second sub-abnormal samples, obtain selected data values.

[0126] In some optional embodiments, in the above-mentioned device 500, the dynamic adjustment coefficient includes: a hard threshold indicator and a soft threshold indicator; the hard threshold indicator is an empirical indicator; the soft threshold indicator is an adaptive indicator obtained by inputting the statistical indicators of the data values ​​in the second sample subset into the regression model for prediction; wherein, the regression model corresponds to the type of converter production data, and the regression model is trained using the statistical indicators of the original data values ​​of each type of marked converter production data as training samples.

[0127] In some optional embodiments, in the above-mentioned device 500, the generation unit 530 is specifically used to: respectively calculate the Pearson correlation coefficient between any two types of converter production data; integrate the Pearson correlation coefficient into a correlation coefficient matrix; convert the correlation coefficient matrix into a color matrix; and form a heat map based on the color matrix.

[0128] The specific definition of the converter production data processing device can refer to the definition of the converter production data processing method mentioned above, which will not be repeated here. Each module in the above converter production data processing device can be implemented in whole or in part by software, hardware and a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0129] In one embodiment, a computer device is provided, whose internal structure diagram can be as follows: Figure 6 As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, the functions or steps of a method for processing converter production data are implemented.

[0130] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program: obtaining data values ​​of a plurality of converter production data for each of a plurality of converters; For data values, a first sample set is obtained through data governance, a first abnormal sample in the first sample set is eliminated using a classification model based on a decision tree to obtain a second sample set, and an adaptive box plot is used to eliminate the second abnormal sample in the second sample set to obtain selected data values ​​of various converter production data; Based on the selected data values, the correlation coefficients between various converter production data are determined, and the correlation coefficients are mapped to the color space to obtain a heat map.

[0131] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: obtaining data values ​​of a plurality of converter production data for each of a plurality of converters; For data values, a first sample set is obtained through data governance, a first abnormal sample in the first sample set is eliminated using a classification model based on a decision tree to obtain a second sample set, and an adaptive box plot is used to eliminate the second abnormal sample in the second sample set to obtain selected data values ​​of various converter production data; Based on the selected data values, the correlation coefficients between various converter production data are determined, and the correlation coefficients are mapped to the color space to obtain a heat map.

[0132] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant description in the aforementioned method embodiment. In order to avoid repetition, they will not be described one by one here.

[0133] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0134] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0135] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for processing converter production data, characterized in that: The method comprises: obtaining data values ​​of a plurality of converter production data of each of the plurality of converters; For the data values, a first sample set is obtained through data governance, a first abnormal sample in the first sample set is eliminated using a classification model based on a decision tree to obtain a second sample set, and a second abnormal sample in the second sample set is eliminated using an adaptive box plot to obtain selected data values ​​of a plurality of the converter production data; According to the selected data values, correlation coefficients between the various converter production data are determined, and the correlation coefficients are mapped to a color space to obtain a heat map.

2. The method for processing converter production data according to claim 1, characterized in that: The step of obtaining data values ​​of a plurality of converter production data of each of the plurality of converters comprises: Using a converter sensor to collect real-time values ​​of a plurality of converter production data of each of the plurality of converters; Persistently storing the real-time value in a database; The data value to be analyzed is retrieved from the database.

3. The method for processing converter production data according to claim 1, characterized in that: The obtaining of the first sample set through data governance includes: Eliminate the data values ​​that are not integer or floating point; Unify the units of the data values; Eliminate the data values ​​exceeding a normal threshold, wherein the normal threshold corresponds to the type of the converter production data, and the normal threshold is determined according to the mean, standard deviation and control threshold sensitivity coefficient of the data value of each type of the converter production data; The retained data values ​​are used as the first sample set.

4. The method for processing converter production data according to claim 1, characterized in that: The step of eliminating the first abnormal sample in the first sample set by using the classification model based on the decision tree to obtain the second sample set includes: Inputting the first sample set into the decision tree-based classification model; Determining the first abnormal sample in the first sample set according to the output of the decision tree-based classification model; After removing the first abnormal samples, the second sample set is obtained; Wherein, the decision tree-based classification model is trained using the marked historical data values ​​of the various converter production data as training samples.

5. The method for processing converter production data according to claim 1, characterized in that: The method of using the adaptive box plot to remove the second abnormal sample in the second sample set to obtain a plurality of selected data values ​​of the converter production data includes: dividing the second sample set into a plurality of second sample subsets according to the types of the converter production data; For each of the second sample subsets, a box is constructed according to the first quartile, the median and the third quartile of the second sample subset, an outlier boundary of the box is determined according to a dynamic adjustment coefficient, and the data values ​​in the second sample subset that are outside the outlier boundary are determined as second sub-abnormal samples; After eliminating a plurality of the second abnormal sub-samples, the selected data value is obtained.

6. The method for processing converter production data according to claim 5, characterized in that: The dynamic adjustment coefficient includes: a hard threshold index and a soft threshold index; The hard threshold indicator is an empirical indicator; The soft threshold index is an adaptive index obtained by inputting the statistical index of the data values ​​in the second sample subset into a regression model for prediction; The regression model corresponds to the type of the converter production data, and the regression model is trained using the statistical indicators of the marked original data values ​​of each type of converter production data as training samples.

7. The method for processing converter production data according to claim 1, characterized in that: Determining the correlation coefficients between the multiple converter production data according to the selected data values, and mapping the correlation coefficients to a color space to obtain a heat map, comprises: Calculate the Pearson correlation coefficient between any two types of converter production data respectively; Integrate the Pearson correlation coefficients into a correlation coefficient matrix; Converting the correlation coefficient matrix into a color matrix; The heat map is formed based on the color matrix.

8. A converter production data processing device, characterized in that: The device comprises: An acquisition unit, used for acquiring data values ​​of a plurality of converter production data of each of the plurality of converters; A processing unit is used to obtain a first sample set by data governance for the data values, eliminate first abnormal samples in the first sample set by using a classification model based on a decision tree to obtain a second sample set, and eliminate second abnormal samples in the second sample set by using an adaptive box plot to obtain selected data values ​​of a plurality of converter production data; A generating unit is used to determine the correlation coefficients between the multiple converter production data according to the selected data values, and map the correlation coefficients to a color space to obtain a heat map.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for processing converter production data according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for processing converter production data according to any one of claims 1 to 7 are implemented.

Images