System and method for predicting basicity of slag in silicon-manganese alloy smelting

Abstract

The present invention provides a system and method for predicting the basicity of slag in a silicon-manganese alloy smelting process. The system comprises: a data acquisition module for acquiring chemical analysis data of the raw material ore, the ratio of the raw material ore to the batching material, and chemical analysis data of the silicon-manganese alloy and the slag after smelting; a data calculation module for calculating the chemical composition of the batching material and product indicators; a model training module for training a neural network model based on the calculated values ​​to obtain a prediction model; a data preprocessing module for mapping and reconstructing the batching material and product data with lags through time cross-correlation analysis, and extracting characteristic values ​​of the batching material and product data through feature engineering; and a prediction module for inputting the characteristic values ​​into the prediction model to predict the basicity of the slag at a future time. The present invention fully considers the time delay between the time of material addition and the time of smelting completion. The model trained using the processed data accurately reflects the relationship between the batching material and the product, and has high calculation accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of silico-manganese slag composition prediction, and in particular to a system and method for predicting the basicity of silico-manganese alloy smelting slag. Background Art

[0002] The composition of silicon-manganese slag is an important factor affecting the technical and economic indicators of smelting. It is very important to select the appropriate slag composition according to the raw materials and equipment conditions. Although the conditions for smelting silicon-manganese alloys are different, the general principle is to have a suitable melting point, viscosity and conductivity to ensure good technical and economic indicators. Among them, slag basicity is the most important indicator of slag composition. Appropriate basicity is one of the important conditions for improving manganese recovery rate, increasing output and reducing smelting power consumption. The definition of slag basicity is as follows:

[0003]

[0004] Where R3 is the ternary basicity of the slag; CaO is the percentage of calcium oxide in the slag; MgO is the percentage of magnesium oxide in the slag; and SiO2 is the percentage of silicon dioxide in the slag. It is generally believed that the basicity of slag should be controlled between 0.55 and 0.65.

[0005] The traditional method for adjusting slag basicity involves testing the slag after one furnace of silicon-manganese alloy smelting is completed. Based on the calculated slag basicity, the ratio of raw materials such as manganese ore, dolomite, silica, and coke in the next furnace is adjusted. Generally, after three to four furnaces of adjustment, the results and characteristics of the slag basicity adjustment can be confirmed. Because each furnace of silicon-manganese alloy smelting takes three to four hours, the adjustment of slag basicity has a significant lag.

[0006] Patent document CN103361461A (application number: CN201210092210.9) discloses an online prediction and control method for phosphorus content in converter-smelted low-carbon steel. Its characteristics are: (1) using online collected furnace gas information to indirectly predict the phosphorus content in molten steel online in real time; (2) using the indirect relationship between the change of CO in the converter gas and the oxygen content in the molten pool in the late blowing stage, that is, predicting the carbon content from the change of CO in the gas, and then predicting the oxygen content based on the carbon-oxygen product relationship, and controlling the slag basicity at R = 3.0-3.5; (3) judging the feeding amount, oxygen consumption and blowing end point, combining mathematical models and operating processes to achieve online control of phosphorus content in converter-smelted low-carbon steel. However, this patent cannot predict the change trend and range of slag basicity in advance before determining the raw material ore to be fed into the submerged arc furnace, and cannot optimize the operation of the submerged arc furnace by determining the adjustment amount of the raw material and auxiliary material ratio in advance, solve the problem of untimely feedback, and ultimately improve the output of the submerged arc furnace and the quality of silicon-manganese alloy. Summary of the Invention

[0007] In view of the defects in the prior art, the purpose of the present invention is to provide a system and method for predicting the basicity of silicon-manganese alloy smelting slag.

[0008] The system for predicting the basicity of silicon-manganese alloy smelting slag provided by the present invention comprises:

[0009] Data acquisition module: acquires the test data of raw ore, the ratio data of raw ore and ingredients, and the test data of silicon manganese alloy and slag after smelting;

[0010] Data calculation module: calculate the chemical composition of ingredients and product indicators;

[0011] Model training module: trains the neural network model according to the calculated values ​​to obtain a prediction model;

[0012] Data preprocessing module: This module maps and reconstructs the lagged ingredient data and product data through time cross-correlation analysis, and extracts the characteristic values ​​of the ingredient data and product data through feature engineering;

[0013] Prediction module: Input the characteristic value into the prediction model to predict the slag basicity at future moments.

[0014] Preferably, the data acquisition module includes:

[0015] Obtain the ratio of raw ore and auxiliary materials for each batching, calculate the comprehensive moisture content, manganese content, iron content, silicon content, calcium content, magnesium content, aluminum content, manganese-iron ratio, silicon-manganese ratio, carbon-manganese ratio, aluminum-manganese ratio of the input components, and the theoretical basicity of the raw materials entering the furnace;

[0016] Obtain the furnace data of each completed silicon-manganese alloy smelting, including the start and end time of smelting, the output of silicon-manganese alloy, the amount of slag, the manganese content in the silicon-manganese alloy, the silicon content, the manganese content in the slag, the silicon dioxide content, the magnesium oxide content, the calcium oxide content, and the calculation of the slag basicity, as well as the overall content of manganese, iron, silicon, calcium, magnesium, and aluminum in the silicon-manganese alloy and the slag.

[0017] Preferably, in the ingredient data and product data, manganese content, iron content, calcium content, magnesium content, and alkalinity indicators are selected, and the time correlation coefficient of each indicator is calculated with a step length of 15 minutes;

[0018] Add up the correlation coefficients of all indicators under the same time step, and the time step corresponding to the maximum result is the delay time between feeding and unloading;

[0019] The timestamps corresponding to all indicators are shifted backward according to the calculated delay time, mapped one-to-one with the timestamps of the furnace output indicators, and the two data sets are spliced ​​to obtain the corresponding data sets of the feeding and product indicators.

[0020] Preferably, the Pearson correlation coefficients of slag basicity and other indicators are calculated respectively, and the correlation coefficients are arranged from large to small, and the variables with correlation coefficients higher than 0.5 are selected as the characteristic values ​​of the prediction model;

[0021] Construct the data set L based on the slag basicity and characteristic values;

[0022]

[0023] Where y is the slag basicity, x is the slag basicity-related characteristic value obtained through correlation analysis, and the subscripts a, b, c, n, and t are all natural numbers greater than 0.

[0024] Preferably, a neural network prediction model is constructed, and the model structure includes three fully connected layers, wherein the first and second layers use the Relu function as the activation function, and the third layer uses the linear function as the activation function. A Dropout layer is added between each layer to enhance the generalization ability of the prediction model;

[0025] The y in the data set L is used as the output of the neural network model, and the values ​​other than y are used as input. After training, a prediction model for slag basicity is obtained.

[0026] The method for predicting the basicity of silicon-manganese alloy smelting slag provided by the present invention comprises:

[0027] Data acquisition steps: obtaining the test data of the raw material ore, the ratio data of the raw material ore and the ingredients, and the test data of the silicon manganese alloy and slag after the smelting is completed;

[0028] Data calculation steps: Calculate the chemical composition of ingredients and product indicators;

[0029] Model training steps: Train the neural network model based on the calculated values ​​to obtain a prediction model;

[0030] Data preprocessing step: The lagged ingredient data and product data are mapped and reconstructed through time cross-correlation analysis, and the characteristic values ​​of the ingredient data and product data are extracted through feature engineering;

[0031] Prediction step: Input the characteristic value into the prediction model to predict the slag basicity at the future moment.

[0032] Preferably, the data acquisition step includes:

[0033] Obtain the ratio of raw ore and auxiliary materials for each batching, calculate the comprehensive moisture content, manganese content, iron content, silicon content, calcium content, magnesium content, aluminum content, manganese-iron ratio, silicon-manganese ratio, carbon-manganese ratio, aluminum-manganese ratio of the input components, and the theoretical basicity of the raw materials entering the furnace;

[0034] Obtain the furnace data of each completed silicon-manganese alloy smelting, including the start and end time of smelting, the output of silicon-manganese alloy, the amount of slag, the manganese content in the silicon-manganese alloy, the silicon content, the manganese content in the slag, the silicon dioxide content, the magnesium oxide content, the calcium oxide content, and the calculation of the slag basicity, as well as the overall content of manganese, iron, silicon, calcium, magnesium, and aluminum in the silicon-manganese alloy and the slag.

[0035] Preferably, in the ingredient data and product data, manganese content, iron content, calcium content, magnesium content, and alkalinity indicators are selected, and the time correlation coefficient of each indicator is calculated with a step length of 15 minutes;

[0036] Add up the correlation coefficients of all indicators under the same time step, and the time step corresponding to the maximum result is the delay time between feeding and unloading;

[0037] The timestamps corresponding to all indicators are shifted backward according to the calculated delay time, mapped one-to-one with the timestamps of the furnace output indicators, and the two data sets are spliced ​​to obtain the corresponding data sets of the feeding and product indicators.

[0038] Preferably, the Pearson correlation coefficients of slag basicity and other indicators are calculated respectively, and the correlation coefficients are arranged from large to small, and the variables with correlation coefficients higher than 0.5 are selected as the characteristic values ​​of the prediction model;

[0039] Construct the data set L based on the slag basicity and characteristic values;

[0040]

[0041] Where y is the slag basicity, x is the slag basicity-related characteristic value obtained through correlation analysis, and the subscripts a, b, c, n, and t are all natural numbers greater than 0.

[0042] Preferably, a neural network prediction model is constructed, and the model structure includes three fully connected layers, wherein the first and second layers use the Relu function as the activation function, and the third layer uses the linear function as the activation function. A Dropout layer is added between each layer to enhance the generalization ability of the prediction model;

[0043] The y in the data set L is used as the output of the neural network model, and the values ​​other than y are used as input. After training, a prediction model for slag basicity is obtained.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] (1) The data acquisition and processing of the present invention fully considers the time delay between the feeding time and the smelting end time. The model trained with the processed data can correctly reflect the relationship between the ingredients and the product, and the calculation accuracy is high;

[0046] (2) The present invention predicts the slag alkalinity based on a neural network model, and predicts the slag alkalinity of future products according to the calculation results of the ingredients, so that the alkalinity can be regulated in advance to achieve the purpose of optimizing the smelting process; at the same time, through active regulation, it can effectively avoid large fluctuations in the smelting furnace conditions and product quality, thereby achieving cost reduction and efficiency improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0048] Figure 1 Flow chart of the method of the present invention;

[0049] Figure 2 This is a comparison chart of the predicted value and actual measured value of the slag basicity of the 370 furnace product according to the present invention. DETAILED DESCRIPTION

[0050] The present invention will be described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those skilled in the art, several changes and improvements can be made without departing from the scope of the present invention. These all fall within the scope of protection of the present invention.

[0051] Example:

[0052] like Figure 1 The present invention provides a method for establishing a prediction model for the basicity of silicon-manganese alloy smelting slag, comprising the following steps:

[0053] Step 1: Collect the proportions of various raw materials and auxiliary materials for each batching, calculate the comprehensive moisture content, manganese content, iron content, silicon content, calcium content, magnesium content, aluminum content, manganese-iron ratio, silicon-manganese ratio, carbon-manganese ratio, aluminum-manganese ratio of the input components, and the theoretical basicity of the raw materials entering the furnace;

[0054] Step 2, collecting the furnace discharge data of each silicon-manganese alloy smelting completion, including the start and end time of smelting, the output of silicon-manganese alloy, the amount of slag, the manganese content in the silicon-manganese alloy, the silicon content, the manganese content in the slag, the silicon dioxide content, the magnesium oxide content, the calcium oxide content, and the calculation of the slag basicity, the total content of manganese, iron, silicon, calcium, magnesium, and aluminum in the silicon-manganese alloy and the slag;

[0055] Step 3: Select the manganese content, iron content, calcium content, magnesium content, and alkalinity indicators from the ingredient data in step 1 and the product data in step 2, and calculate the time correlation coefficient of each indicator with a step length of 15 minutes; add the cross-correlation coefficients of all indicators under the same time step length, and the time step corresponding to the maximum result is the delay time between charging and unloading;

[0056] Step 4: Shift the timestamps corresponding to all indicators in step 1 backward according to the delay time calculated in step 3, map them one-to-one with the timestamps of the furnace-out indicators in step 2, and then concatenate the two data sets to obtain a data set corresponding to the feed and product indicators.

[0057] Step 5: Calculate the Pearson correlation coefficients of slag basicity and other indicators respectively, sort the correlation coefficients from large to small, and select the variables with correlation coefficients higher than 0.5 as the characteristic values ​​of the prediction model;

[0058] Step 6, extracting the slag basicity and the eigenvalues ​​selected in step 5 to construct a data set L;

[0059]

[0060] Where y is the slag basicity, and x is the slag basicity-related characteristic value obtained through correlation analysis.

[0061] Step 7: Construct a neural network prediction model. The model structure includes three Dense layers, where the first and second layers use the Relu function as the activation function, and the third layer uses the linear function as the activation function. A Dropout layer is added between each layer to enhance the generalization ability of the prediction model. The y in the dataset L is used as the output of the neural network model, and the values ​​other than y are used as input. After training, the prediction model of slag basicity can be obtained.

[0062] The present invention also provides a system for establishing a basicity prediction model for silicon-manganese alloy smelting slag, comprising:

[0063] The data acquisition module is used to collect the test data of various raw materials, the ratio data of various raw materials and auxiliary materials each time, and the test data of silicon manganese alloy and slag after smelting;

[0064] Data calculation module, used to calculate the chemical composition of ingredients and various indicators of products;

[0065] The data preprocessing module reconstructs the lagged ingredient data and product data through time cross-correlation analysis and mapping, and extracts the input feature values ​​of the prediction model through feature engineering;

[0066] The model training module is used to train the neural network model according to the values ​​obtained by the data calculation module to obtain a prediction model;

[0067] The prediction module is used to input the data values ​​obtained by the data calculation module into the prediction model to predict the slag basicity at the future moment, such as Figure 2 , which is a comparison chart of the predicted value and the actual measured value of slag basicity.

[0068] Those skilled in the art will appreciate that, in addition to implementing the system, device, and various modules provided by the present invention in purely computer-readable program code, it is entirely possible to implement the same program in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, embedded microcontrollers, and the like by logically programming the method steps. Therefore, the system, device, and various modules provided by the present invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; the modules for implementing various functions can also be considered both software programs for implementing the method and structures within the hardware component.

[0069] The above describes specific embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. The embodiments of this application and the features in the embodiments may be combined with each other in any manner unless there is a conflict.

Claims

1. A system for predicting the basicity of slag in silicon-manganese alloy smelting, characterized in that: include: Data acquisition module: acquires the test data of raw ore, the ratio data of raw ore and ingredients, and the test data of silicon manganese alloy and slag after smelting; Data calculation module: calculate the chemical composition of ingredients and product indicators; Model training module: trains the neural network model according to the calculated values ​​to obtain a prediction model; Data preprocessing module: This module maps and reconstructs the lagged ingredient data and product data through time cross-correlation analysis, and extracts the characteristic values ​​of the ingredient data and product data through feature engineering; Prediction module: inputs the characteristic value into the prediction model to predict the slag basicity at the future moment; In the ingredient data and product data, select the indicators of manganese content, iron content, calcium content, magnesium content, and alkalinity, and calculate the time correlation coefficient of each indicator with a step length of 15 minutes; Add up the correlation coefficients of all indicators under the same time step, and the time step corresponding to the maximum result is the delay time between feeding and unloading; The timestamps corresponding to all indicators are shifted backward according to the calculated delay time, mapped one-to-one with the timestamps of the furnace output indicators, and the two data sets are spliced ​​to obtain the corresponding data sets of the feeding and product indicators.

2. The silicon-manganese alloy smelting slag basicity prediction system according to claim 1, characterized in that: The data acquisition module includes: Obtain the ratio of raw ore and auxiliary materials for each batching, calculate the comprehensive moisture content, manganese content, iron content, silicon content, calcium content, magnesium content, aluminum content, manganese-iron ratio, silicon-manganese ratio, carbon-manganese ratio, aluminum-manganese ratio of the input components, and the theoretical basicity of the raw materials entering the furnace; Obtain the furnace data of each completed silicon-manganese alloy smelting, including the start and end time of smelting, the output of silicon-manganese alloy, the amount of slag, the manganese content in the silicon-manganese alloy, the silicon content, the manganese content in the slag, the silicon dioxide content, the magnesium oxide content, the calcium oxide content, and the calculation of the slag basicity, as well as the overall content of manganese, iron, silicon, calcium, magnesium, and aluminum in the silicon-manganese alloy and the slag.

3. The silicon-manganese alloy smelting slag basicity prediction system according to claim 1, characterized in that: The Pearson correlation coefficients between slag basicity and other indicators were calculated respectively, and the correlation coefficients were arranged from large to small. The variables with correlation coefficients higher than 0.5 were selected as the characteristic values ​​of the prediction model. Construct the data set L based on the slag basicity and characteristic values; Where y is the slag basicity, x is the slag basicity-related characteristic value obtained through correlation analysis, and the subscripts a, b, c, n, and t are all natural numbers greater than 0.

4. The silicon-manganese alloy smelting slag basicity prediction system according to claim 3, characterized in that: Construct a neural network prediction model. The model structure includes three fully connected layers. The first and second layers use the Relu function as the activation function, and the third layer uses the linear function as the activation function. Dropout layers are added between each layer to enhance the generalization ability of the prediction model. The y in the data set L is used as the output of the neural network model, and the values ​​other than y are used as input. After training, a prediction model for slag basicity is obtained.

5. A method for predicting the basicity of silicon-manganese alloy smelting slag, characterized in that: include: Data acquisition steps: obtaining the test data of the raw material ore, the ratio data of the raw material ore and the ingredients, and the test data of the silicon manganese alloy and slag after the smelting is completed; Data calculation steps: Calculate the chemical composition of ingredients and product indicators; Model training steps: Train the neural network model based on the calculated values ​​to obtain a prediction model; Data preprocessing step: The lagged ingredient data and product data are mapped and reconstructed through time cross-correlation analysis, and the characteristic values ​​of the ingredient data and product data are extracted through feature engineering; Prediction step: Input the characteristic value into the prediction model to predict the slag basicity at the future moment; In the ingredient data and product data, select the indicators of manganese content, iron content, calcium content, magnesium content, and alkalinity, and calculate the time correlation coefficient of each indicator with a step length of 15 minutes; Add up the correlation coefficients of all indicators under the same time step, and the time step corresponding to the maximum result is the delay time between feeding and unloading; The timestamps corresponding to all indicators are shifted backward according to the calculated delay time, mapped one-to-one with the timestamps of the furnace output indicators, and the two data sets are spliced ​​to obtain the corresponding data sets of the feeding and product indicators.

6. The method for predicting basicity of silicon-manganese alloy smelting slag according to claim 5, characterized in that: The data acquisition step includes: Obtain the ratio of raw ore and auxiliary materials for each batching, calculate the comprehensive moisture content, manganese content, iron content, silicon content, calcium content, magnesium content, aluminum content, manganese-iron ratio, silicon-manganese ratio, carbon-manganese ratio, aluminum-manganese ratio of the input components, and the theoretical basicity of the raw materials entering the furnace; Obtain the furnace data of each completed silicon-manganese alloy smelting, including the start and end time of smelting, the output of silicon-manganese alloy, the amount of slag, the manganese content in the silicon-manganese alloy, the silicon content, the manganese content in the slag, the silicon dioxide content, the magnesium oxide content, the calcium oxide content, and the calculation of the slag basicity, as well as the overall content of manganese, iron, silicon, calcium, magnesium, and aluminum in the silicon-manganese alloy and the slag.

7. The method for predicting basicity of silicon-manganese alloy smelting slag according to claim 5, characterized in that: The Pearson correlation coefficients between slag basicity and other indicators were calculated respectively, and the correlation coefficients were arranged from large to small. The variables with correlation coefficients higher than 0.5 were selected as the characteristic values ​​of the prediction model. Construct the data set L based on the slag basicity and characteristic values; Where y is the slag basicity, x is the slag basicity-related characteristic value obtained through correlation analysis, and the subscripts a, b, c, n, and t are all natural numbers greater than 0.

8. The method for predicting basicity of silicon-manganese alloy smelting slag according to claim 7, characterized in that: Construct a neural network prediction model. The model structure includes three fully connected layers. The first and second layers use the Relu function as the activation function, and the third layer uses the linear function as the activation function. Dropout layers are added between each layer to enhance the generalization ability of the prediction model. The y in the data set L is used as the output of the neural network model, and the values ​​other than y are used as input. After training, a prediction model for slag basicity is obtained.

Images