Disclosure of Invention
The invention solves the problem that the converter blowing process in the prior art depends on more manual experience, and provides the artificial intelligence-based converter carbon content index evaluation and temperature analysis method which has high practicability and accurate tapping temperature.
The technical scheme of the invention is that the method for evaluating and analyzing the carbon content index of the converter based on artificial intelligence comprises the following steps: the method comprises the following steps:
firstly, acquiring flame images of the converter mouth in real time through a CCD camera which is arranged in front of a converter and is additionally provided with high temperature and dust protection, and shooting flame photos of the converter mouth in different converting periods on site by using a camera;
step two, processing a furnace mouth flame image, extracting furnace mouth flame characteristics, wherein the flame characteristics comprise flame duty ratio and flame brightness, constructing a flame area correction evaluation model, establishing a preliminary estimation model of a converter carbon content index according to the flame characteristics, and evaluating the converter carbon content index;
step three, establishing a carbon content index adjustment model as a final carbon content index evaluation model to evaluate the carbon content index of the molten steel;
step four, a carbon content index sequence in the converting process can be obtained based on a final carbon content evaluation model, the corresponding relation between the carbon content index of molten steel and a converting end point is analyzed according to the carbon content index sequence, the converting end point of the converter is judged, and when the converter steelmaking reaches the converting end point, the oxygen blowing into the converter is stopped by the oxygen gun oxygen supply system of the converter;
fifthly, presetting a tapping range of a certain steel grade as [ T1, T2], and if the system detects that the temperature of molten steel at the end point of the converter is within the tapping temperature range, meeting tapping conditions, tilting the furnace body, pouring the molten steel into a ladle, and performing a subsequent cast steel process;
and step six, if the temperature of the molten steel of the converter does not meet the tapping temperature, constructing different temperature analysis models, and adjusting the temperature of the molten steel.
In the first step, the camera acquires working images of converter blowing every 5S, wherein the frame rate of the camera is 20 frames per second.
The second step comprises the following steps:
step 2.1, the specific process of calculating the flame ratio is as follows: firstly, adding pixels with pixel values of 1 in the image, wherein the sum of the pixels is the flame area in the image and is marked as S c C=1, 2 … N, c is the image frame, S c Representing the flame area in the image of the c frame, the flame area correction model specifically comprises:
wherein O is 0 The average oxygen content in the whole image acquisition time, wherein O is the real-time oxygen content; y is Y 0 The maximum furnace age of the converter is obtained according to experience statistics, and Y is the actual furnace age of the converter when detecting flame images;
step 2.2, constructing a flame duty ratio analysis model for calculating the flame duty ratio, wherein the flame duty ratio analysis model function is as follows:
wherein m and n are the sizes of the images, S' c For model estimated flame area τ c The flame duty ratio in the selected frame c image is selected; taking the flame binary image as the ROI area and multiplying the ROI area with the original RGB imageThe method is operated to obtain an RGB image only containing flame, finally, the brightness information of the flame is calculated, and after the corresponding RGB flame image is obtained, the flame brightness analysis method specifically comprises the following steps: the values of R, G and B of the RGB image are normalized:
and then obtaining the maximum value of R ', G ', B ' after normalization, wherein the obtained maximum value is the brightness value of the flame image:
Cmax=max(R’,G’,B’)
V i,j =Cmax
step 2.2, constructing a flame brightness analysis model according to the brightness information of the image:
wherein ρ is c Representing the brightness of the flame in the image of the c frame, the greater the function value of the model, the higher the brightness of the flame.
Setting alpha and beta as the ratio of flame to the ratio of brightness of the flame to the carbon content index, wherein the ratio is as follows: α=0.4, β=0.6; the carbon content index estimation pre-model expression is:
ω c =exp(ατ c +βρ c )+A
wherein A is an adjustable parameter of a model, a carbon content index adjustment model is established based on analysis and comparison of converter temperature and tapping temperature, the tapping range of steel types is set to be [ T1, T2], the temperature in the converter is analyzed, and a corresponding carbon content adjustment model is constructed based on the converter temperature, so that an accurate carbon content index is obtained; when the temperature of the converter is lower than the minimum value of the optimal temperature range, namely T < T1, the temperature in the converter is too low, so that the predicted value of the carbon content index is lower, a carbon content index adjustment function is constructed to ensure the evaluation accuracy of the final carbon content index, and the final evaluation model expression of the carbon content index is as follows:
ω' c =ω c ln(5ΔT'+e)
if the temperature in the converter is T2 is less than or equal to T1, the temperature is less than or equal to TAs the optimal converter temperature, analyzing the converter temperature, and constructing a carbon content index adjustment model, wherein the final evaluation model of the carbon content index in the temperature range is as follows:
wherein r is a conversion factor, and is set to 0.5; when the temperature in the converter is higher than the maximum value of the optimal temperature range, namely T > T2, the temperature in the converter is considered to be too high, and the final carbon content index evaluation model expression is as follows:
the final carbon content evaluation model established above was for:
wherein ω' c Representing the most accurate calculation of the carbon content indicator.
In the fourth step, the carbon content index calculation is performed on a series of acquired image data to obtain a carbon content index sequence: [ omega ] 1 ,ω 2 …ω N ]According to the obtained carbon contentAnalyzing and judging the converting end point of the converter steelmaking by the quantitative index sequence; assuming that the index range of the carbon content when a certain steel product reaches the converting end point is [ omega 1, omega 2]]The method comprises the steps of carrying out a first treatment on the surface of the Calculating the carbon-containing index of the converter according to the final carbon content index evaluation model, wherein when the carbon-containing index obtained by continuous multiple N frames is in [ omega 1, omega 2]]When the converter steelmaking reaches the converting end point, the system sends out an instruction to the oxygen supply system of the converter oxygen lance, stops blowing oxygen into the converter, and sets the carbon-containing index acquired in continuous 5 frames to be [ omega 1, omega 2]]And if the carbon content index does not change greatly within the range and after 10 continuous frames, the converting endpoint is considered to be reached.
The temperature analysis model meeting the tapping condition in the fifth step is as follows: firstly judging a converter endpoint according to a carbon content index, analyzing the temperature of molten steel in a converter when the converter endpoint is reached, comparing the temperature of the molten steel in the converter with the tapping temperature range of steel types, setting the tapping range of the steel types to be [ T1, T2], and if the system detects that the temperature of the converter is not less than T2 and not more than T1 in the tapping temperature range, satisfying the tapping condition, tilting the converter body, and pouring the molten steel into a ladle for subsequent cast steel procedures.
The specific process of the temperature analysis model which does not meet the tapping condition in the step six is as follows:
when T < T1, the temperature of molten steel in the converter is too low after the converter reaches the converting end point, if the molten steel is poured into a ladle, the pouring operation cannot be smoothly performed, the quality of the steel is affected, the pouring cannot be caused when the molten steel is serious, a molten steel temperature raising analysis model is built, and the temperature of the molten steel in the converter is raised, wherein the temperature raising analysis model expression is as follows:
wherein k is an adjustable coefficient, m is the mass of the temperature raising agent, the unit is kg, the value of the adjustable coefficient of the model can be determined according to the type of the selected temperature raising agent by an implementer according to different values of the temperature raising agent k, and the value is set to be 2;
when T > T2, the molten steel temperature is too high, the molten steel is easy to suck, the converter age and the alloy yield are reduced, in order to reduce the molten steel temperature and prevent the burning accident caused by the molten steel temperature being too high, a function model for reducing the molten steel temperature is constructed, and the cooling model function is specifically as follows:
M=R(T-T2)
wherein M is the amount of coolant added in the converter, the unit is kg, R is the adjustment factor, and r=16 is set.
Compared with the prior art, the artificial intelligence-based converter carbon content index evaluation and temperature analysis method has the following advantages: firstly, a converter carbon content index estimation model is established according to the flame characteristics of a furnace mouth, different carbon content index adjustment models are established for preventing phenomena of lower carbon pulling, higher carbon pulling and the like caused by inaccurate carbon content index prediction, a final carbon content index estimation model is obtained according to different conditions, the error value of a system for estimating the carbon content index of the converter is reduced, a steelmaking endpoint is accurately identified further based on the carbon content index of molten steel, the accurate judgment of the system on the smelting endpoint is improved, the endpoint hit rate is high, the judgment result is accurate and reliable, the working efficiency of the system is improved, the real-time performance is realized, and the uncertain factors of manual identification are reduced.
Meanwhile, the problems that pouring operation is difficult, steel quality is reduced and the like caused by too high and too low temperature during molten steel tapping are considered, the temperature of the molten steel of the converter after stopping blowing is compared with the tapping temperature range, and a temperature analysis model is built, so that the temperature of the molten steel of the converter can meet the tapping temperature range, and the problems of steel product quality reduction, safety accidents and the like caused by uncomfortable tapping temperature are prevented.
The system has higher practicability, can automatically acquire the carbon content index of the molten steel and accurately judge the converting end point of the converter, and simultaneously adjusts the temperature of the molten steel according to the established temperature analysis model so as to ensure that the tapping temperature range is met.
Detailed Description
The method for evaluating the carbon content index and analyzing the temperature of the converter based on artificial intelligence is further described below with reference to the accompanying drawings and the specific embodiments:
as shown in the figure, in the embodiment, a converter carbon content index evaluation model is firstly established according to the flame characteristics of a furnace mouth, and the influence of the temperature in the furnace on the judgment of the carbon content index is considered, so that a final carbon content index evaluation model is established for accurately evaluating the carbon content index of molten steel. And acquiring a carbon content index sequence in the converting process according to the final carbon content index evaluation model, and judging the converting end point of the converter. When the converter steelmaking reaches the converting end point, the system sends an end point control instruction, the oxygen gun oxygen supply system of the converter is closed, and oxygen blowing into the converter is stopped. Meanwhile, the system adopts a thermocouple temperature measuring device to measure the temperature of a molten pool, so that the temperature of molten steel meets the tapping temperature condition, the quality of steel is ensured, different models are established aiming at the condition that the temperature in a furnace is too high and too low, and the temperature of the molten steel is analyzed and regulated so as to ensure that the temperature of the molten steel can meet the tapping temperature range when the blowing process reaches the end point, and the problems of steel product quality reduction, safety accidents and the like caused by the fact that the tapping temperature is too high and too low are avoided.
The specific main implementation steps are as follows:
firstly, acquiring flame images of a converter mouth of a converter in real time through a CCD camera which is arranged in front of the converter and is added with high temperature and dust protection, and shooting flame photos of the converter mouth in different converting periods on site by using a camera;
step two, processing a furnace mouth flame image, extracting furnace mouth flame characteristics, wherein the flame characteristics comprise flame duty ratio and flame brightness, constructing a flame area correction evaluation model to ensure that accurate flame duty ratio is acquired later, and finally, establishing a preliminary estimation model of a converter carbon content index according to the flame characteristics for evaluating the converter carbon content index;
step three, taking the fact that the judgment of the carbon content is error due to the fact that the temperature is too high or too low into consideration, establishing a carbon content index adjustment model as a final carbon content index evaluation model to accurately estimate the carbon content index of molten steel;
step four, a carbon content index sequence in the converting process can be obtained based on a final carbon content evaluation model, the corresponding relation between the carbon content index of molten steel and a converting end point is analyzed according to the carbon content index sequence, the converting end point of the converter is judged, and when the converter steelmaking reaches the converting end point, the oxygen blowing into the converter is stopped by the oxygen supply system of the oxygen lance of the converter;
fifthly, tapping can be performed when the tapping temperature of the molten steel meets the tapping temperature range for different steel types. Presetting a tapping range of a certain steel grade as [ T1,2], if the system detects that the temperature of molten steel at the end point of a converter is within the tapping temperature range, meeting tapping conditions, tilting a furnace body, pouring the molten steel into a ladle, and performing a subsequent cast steel process;
step six, if the temperature of the molten steel of the converter does not meet the tapping temperature, constructing different temperature analysis models, and adjusting the temperature of the molten steel to ensure that the temperature of the molten steel reaching the end point is within the tapping temperature range of the steel product, so as to prevent the problems of steel product quality reduction and the like caused by overhigh and overlow tapping temperature.
The specific detailed implementation steps are as follows:
1. firstly, preliminary rough estimation is carried out on the carbon content index of the converter steelmaking through image features, then, the accurate carbon content index of molten steel is obtained through a final adjustment model, the calculation accuracy of the carbon content index is improved, and then, the temperature of the molten steel which does not meet the tapping temperature is regulated and controlled, so that the temperature of the converter reaching the end point can meet the tapping temperature range of the molten steel.
2. Firstly, a flame image of the converter mouth is acquired through a camera, the camera is a CCD camera which is fixed in front of the converter and is added with high temperature and dust protection, and the flame image of the converter mouth can be acquired in real time so as to analyze the flame subsequently. In order to reduce the power consumption of the camera, the acquisition frame rate of the camera is set to be 20 frames per second, the camera acquires images every 5S, and the camera acquires images to record the whole process of converting of the converter for subsequent analysis of carbon content indexes of a converting end point and the like.
3. The flame at the furnace mouth and the spark at the furnace mouth are an important judging basis of the carbon content of molten steel, the characteristics of the flame are extracted, the carbon content of the molten steel is estimated according to the characteristics, the flame characteristics are mainly extracted, the brightness degree and the flame duty ratio of the flame are used as flame characteristic values in different periods, and the carbon content index of the molten steel is estimated based on the characteristic values.
4. The flame image is segmented to extract the flame area, so as to extract the flame in the image, and a plurality of image segmentation methods exist: a threshold-based segmentation method, region-based image segmentation, an edge-based segmentation algorithm, a wavelet transform-based segmentation method, and the like. A threshold-based image segmentation algorithm is used to extract the flame.
5. The selection of the threshold value in the image segmentation process is determined by adopting a histogram, and the method has the advantages that the contrast between the brightness of the flame and the brightness of the black converter is strong, so that the distribution of background pixels and flame pixels on the image on the histogram can show bimodality, and the gray value corresponding to the valley bottom between the two peaks is used as the segmentation threshold value T' of the image.
In the formula, T 'is a flame segmentation threshold, and the value selection operator needs to determine according to the gray level histogram of the image, and T' =150 is set according to the gray level characteristic of the flame.
6. Thus, a flame binary image can be obtained.
7. Preliminary prediction is carried out on the carbon content index of the converter by extracting the flame duty ratio and the flame brightness characteristics, and the flame duty ratio is calculated firstly, wherein the specific process is as follows:
1) Firstly, adding pixels with pixel values of 1 in the image, wherein the sum of the pixels is the flame area in the image and is marked as S c C=1, 2 … N, c is the image frame, S c Representing the flame area in the c-th frame image. In order to obtain accurate flame area, the flame area obtained based on the image is corrected based on the oxygen content in the converter and the furnace age so as to estimate the accurate flame area in the converter and ensure the evaluation accuracy of the subsequent carbon content index. The flame area correction model specifically comprises the following steps:
wherein O is 0 The average oxygen content in the whole image acquisition time, wherein O is the real-time oxygen content; y is Y 0 The maximum furnace age of the converter is obtained according to experience statistics, and Y is the actual furnace age of the converter when the flame image is detected.
2) Then, a flame duty ratio analysis model is constructed and used for calculating the flame duty ratio, and the flame duty ratio analysis model function is as follows:
wherein m and n are the sizes of the images, S' c For model estimated flame area τ c And the flame duty ratio in the selected frame c image is selected. So far, the flame duty ratio can be obtained and used as the characteristic data for preliminary evaluation of the converter carbon content index judgment later.
8. For flame brightness acquisition: in order to facilitate a subsequent system to accurately obtain the brightness of the flame, taking the flame binary image as an ROI (region of interest) region, performing multiplication operation on the ROI region and an original RGB (red, green and blue) image only containing the flame, and finally calculating the brightness information of the flame.
9. After the corresponding RGB flame image is acquired, the flame brightness analysis method specifically comprises the following steps: the values of R, G and B of the RGB image are normalized:
and then obtaining the maximum value of R ', G ', B ' after normalization, wherein the obtained maximum value is the brightness value of the flame image:
Cmax=max(R′,G’,B’)
V i,j =Cmax
finally, constructing a flame brightness analysis model according to the brightness information of the image:
wherein ρ is c Representing the brightness of the flame in the image of the c frame, the greater the function value of the model, the higher the brightness of the flame.
So far, the brightness of the flame can be calculated, so that a subsequent system can analyze the converter carbon content index based on the characteristic data.
10. With the change of the blowing period, the carbon content is continuously reduced when the blowing period reaches the end of the blowing, the carbon in the steel furnace is exhausted, the chemical reaction tends to be stable, the flame is sparse, the flame is softer, and the texture is fine, so that the carbon content index in the steel furnace is expressed by adopting the flame ratio and the brightness characteristic of the flame.
11. In order to quickly calculate the carbon content index in the converter, a prediction estimation model is firstly constructed according to flame characteristics, so that the carbon content index of molten steel in the converter is initially estimated. Setting alpha and beta as the ratio relation between the flame duty ratio and the brightness degree and the carbon content index of the flame, and setting the value as follows: α=0.4, β=0.6, which can be chosen by the practitioner. The carbon content index estimation pre-model expression is:
ω c =exp(ατ c +βp c )+A
wherein A is an adjustable parameter of the model, and an operator selects the model by himself, so that a preliminary estimation model of the carbon content index can be obtained according to the method.
12. Considering that the temperature in the furnace can influence the prediction of the carbon content, the system establishes a carbon content index adjustment module, and aims to obtain an accurate molten steel carbon content index and prevent the phenomena of lower carbon pulling or higher carbon pulling, and the like caused by inaccurate carbon content estimation from affecting the quality of subsequent steel products. In the whole converter steelmaking process, the temperature requirement is strict, the temperature in the converter needs to be ensured to be in a proper temperature range, and when the temperature of the converter is in the molten steel tapping temperature range, the detected carbon content index is accurate, so that the carbon content index estimation is not accurate enough only based on a flame image, when the temperature of the converter is high, the chemical reaction speed is high, the flame brightness degree is high, the flame area is large, and under the condition that the carbon content index is the same, the carbon content index predicted according to the characteristics of the flame image is higher; when the temperature in the furnace is too low, the carbon oxidization speed is slow, the flame shrinkage is early, the brightness is low, and the estimated carbon content index according to the flame image is lower than the actual value.
13. Aiming at the problems, a carbon content index adjustment model is constructed and used for accurately calculating the carbon content index of the converter under different conditions. The final carbon content index adjustment model is established based on analysis and comparison of the temperature of the converter and the tapping temperature, and a thermocouple temperature measuring device is arranged beside the converter for measuring the temperature in the converter. And measuring the temperature T in the furnace by using thermocouple temperature measuring equipment, performing comparative analysis on the temperature T in the furnace and the tapping temperature of molten steel, constructing a final carbon content index adjustment model according to the temperature difference, setting the tapping range of a certain steel grade as [ T1, T2] for different tapping temperature ranges of different steel grades, analyzing the temperature in the furnace, and constructing a corresponding carbon content adjustment model based on the temperature of the converter so as to obtain an accurate carbon content index. The specific carbon content final evaluation model adjustment method is as follows:
when the temperature of the converter is lower than the minimum value of the optimal temperature range, namely T < T1, the temperature in the converter is too low, so that the predicted value of the carbon content index is lower, and in order to accurately judge the end point of the converter according to the carbon content index later, a carbon content index adjusting function is constructed to ensure the assessment precision of the final carbon content index, the final assessment model expression of the carbon content index is as follows:
ω′ c =ω c ln(5ΔT′+e)
if the temperature in the converter is T2 and T1, the temperature in the converter is considered to be in a proper temperature range,at this time, the carbon content index based on the flame image detection is relatively accurate, but in order to reduce the prediction error caused by the flame image, the accuracy of the subsequent end point judgment is ensuredAs the optimal converter temperature, the converter temperature is analyzed, a carbon content index adjusting model is constructed, and the carbon content index of the converter in a proper temperature range is calculated more accurately, so that a final evaluation model of the carbon content index in the temperature range is set as follows:
wherein r is a conversion factor, which can be selected by an implementer according to practical situations and set to 0.5.
When the temperature in the converter is higher than the maximum value of the optimal temperature range, namely T is more than T2, the temperature in the converter is considered to be too high, a function with larger compensation degree is constructed and used for further correcting the carbon content index of the converter so as to obtain an accurate carbon content index and prevent the problems of end point judgment errors and the like caused by inaccurate evaluation of the carbon content index. The final carbon content index evaluation model expression is obtained as follows:
the final carbon content evaluation model established above was for:
wherein ω' c Representing the most accurate calculation of the carbon content indicator.
14. Therefore, the final carbon content index evaluation model can be used for accurately calculating the carbon content index of the molten steel of the converter, so that the prediction precision of the carbon content index is improved, the blowing endpoint can be further and accurately judged, and meanwhile, the carbon content index of the steel product of the converter can be conveniently known in real time by subsequent operators, and the quality of the steel product is improved.
15. The carbon content index calculation is carried out on a series of acquired image data, and a carbon content index sequence can be obtained: [ omega ] 1 ,ω 2 …ω N ]And analyzing and judging the converting end point of the converter steelmaking according to the obtained carbon content index sequence.
16. The carbon content index of the blowing end point is different for different steel types, and the carbon content index range when a certain steel product reaches the blowing end point is [ omega 1, omega 2]. And calculating a converter carbon-containing index according to the final carbon content index evaluation model, and when the carbon-containing index obtained by continuously obtaining a plurality of N frames is [ omega 1, omega 2], considering that the converter steelmaking reaches a converting end point, and sending an instruction to a converter oxygen lance oxygen supply system by the system at the moment to stop blowing oxygen into the converter. Setting the carbon content index obtained by continuous 5 frames within the range of [ omega 1, omega 2] and setting the carbon content index of continuous 10 frames after the carbon content index does not change greatly, and considering that the converting end point is reached.
17. Since the temperature of the converter molten steel during tapping is a key factor affecting the quality of steel products, in order to obtain high-quality steel products, when the converter reaches the end point, the temperature of the molten steel is analyzed, and different rules are set for analyzing and controlling the temperature of the molten steel which does not meet the tapping temperature.
18. The temperature analysis model specifically comprises the following steps: firstly judging a converter endpoint according to a carbon content index, analyzing the temperature of molten steel in a converter when the converter endpoint is reached, comparing the temperature of the molten steel in the converter with the tapping temperature range of steel types, setting the tapping range of the steel types to be [ T1, T2], and if the system detects that the temperature of the converter is not less than T2 and not more than T1 in the tapping temperature range, meeting the tapping condition, tilting the converter body, and pouring the molten steel into a ladle for subsequent cast steel working procedures. The tapping temperatures are different from one steel grade to another, and the practitioner sets the tapping temperature range according to the type of steel grade.
19. When the converter steelmaking reaches the end point, the converter temperature does not meet the tapping temperature of molten steel, and a specific temperature analysis model is constructed to adjust the molten steel temperature under the condition so as to ensure that the temperature during tapping of the molten steel can reach the standard tapping temperature range. The specific process of the temperature analysis model is as follows:
when T is less than T1, the temperature of molten steel in the converter is too low after the converter reaches the converting end point, if molten steel is poured into a ladle, the pouring operation cannot be smoothly performed, the quality of the steel is affected, the pouring cannot be caused when the quality is serious, a molten steel temperature raising analysis model is built, the temperature of the molten steel in the converter is raised, and the temperature raising analysis model expression is as follows:
wherein k is an adjustable coefficient, m is the mass of the temperature raising agent, and the unit is kg. For different temperature raising agents, the value of the adjustable coefficient of the model can be determined by an implementer according to the type of the selected temperature raising agent, and the value is set to be 2.
When T is more than T2, the molten steel temperature is too high, the molten steel is easy to suck, the converter age and the alloy yield are reduced, in order to reduce the molten steel temperature and prevent the burning accident caused by the molten steel temperature which is too high, a function model for reducing the molten steel temperature is constructed, and the function of the cooling model is specifically as follows:
M=R(T-T2)
wherein M is the addition amount of the coolant in the converter, the unit is kg, R is an adjusting factor, and the value implementation can be selected by the user according to actual conditions, and R=16 is set.
20. Thus, it is completed. The system can accurately and rapidly predict the carbon content index in the converter converting process, judge the converting end point based on the obtained carbon content index sequence, and control the temperature of molten steel according to the temperature analysis model so as to ensure that the tapping temperature can be met when the molten steel is tapped, and prevent the problems of steel product quality, safety accidents and the like caused by the excessively high tapping temperature and the excessively low tapping temperature.