RU2772726C1 - Method for predicting metal emission in a converter, a method for producing liquid steel from molten steel-making iron in a converter, and a system for predicting metal emission in a converter - Google Patents

Abstract

FIELD: metallurgy.

SUBSTANCE: invention relates to metallurgy and can be used to predict the occurrence of metal emissions during decarburizing refining of molten steel-making iron in a converter. The method includes measuring the emission spectrum of the flame from the combustion process in the neck coming out of the neck of the converter, calculating the radiation intensity in the measured radiation spectrum at a wavelength in the range of 580 - 620 nm and predicting the occurrence of a metal emission based on a change in the calculated radiation intensity in a time series, which is determined in accordance with the definition formula, in which a sliding mean is used.

EFFECT: invention makes it possible to blow cast iron with oxygen, preventing the occurrence of metal emission, without the need to install a sensor to detect metal emission inside the furnace, as well as reliably suppress metal emission according to the results of forecasting.

7 cl, 4 dwg, 3 tbl, 3 ex

Description

The technical field to which the invention belongs

The present invention relates to a method and system for predicting the release of metal (the release of slag and liquid iron from a converter) during decarburization refining of molten pig iron in a converter. The present invention also relates to a method for operating a converter, a method for purging molten pig iron with oxygen while preventing the occurrence of a metal blowout.

Prior Art

The molten pig iron discharged from the blast furnace is charged into the converter, and the oxidizing gas (oxygen gas) is supplied from the top blow lance or the bottom blow lance to the molten pig iron loaded into the converter to carry out decarburization refining of liquid iron in the converter, and liquid steel is produced by steel smelting from molten pig iron. In this type of converter, an oxidizing gas is blown into the converter (called "oxygen purge") to dissolve the flux in the slag, resulting in slag, and impurity elements (such as P and Si) contained in the molten pig iron are removed into the slag. However, when attempting to sufficiently dissolve the flux in the slag, this can often result in a so-called "blowout" where the resulting slag causes foaming and the slag and molten iron (molten pig iron or molten steel) in the converter jolt outward through the converter neck during oxygen purge.

There is an opinion that, in particular, when a large amount of iron oxide sources (such as iron ore and mill scale) are added to the converter, or when the converter is operated in soft purge mode, metal is released due to an increase in the accumulation of oxygen (amount of FeO) in the slag. , and the decarburization reaction (C + O → CO) starts violently at the interface between the slag and molten iron (molten pig iron or molten steel), giving a large amount of CO gas.

The ejection of metal unbalances the molten steel components, reduces the output of the output, and therefore causes various problems, such as an increase in the time of decarburization and refining, a decrease in the degree of gas recovery in OG equipment (equipment for the treatment of off-gases without combustion), deterioration in production conditions and output failure of peripheral devices. To deal with these problems, various methods have been proposed in the past for predicting the release of metal.

For example, Patent Literature 1 proposes a method for predicting a metal kick by measuring the top lance vibration with a vibration sensor mounted on the top blast lance, calculating the percentage of signals with an amplitude greater than a given amplitude value among all measured vibration signals at a given time, and determining whether that a metal ejection will occur when the calculated percentage exceeds a predetermined value.

Patent Literature 2 proposes a refining method of directing microwaves to a slag surface in a converter, capturing the microwaves reflected from the slag surface, calculating the mixed wave frequency from the incident wave and the reflected wave and/or the microwave reflection coefficient at the slag surface, determining the level of the slag, and the slag formation progress based on the calculated value(s), and setting and monitoring the corresponding influencing factors so that the slag level and the slag formation progress are maintained at a predetermined standard state.

Patent Literature 3 proposes a converter refining method in a converter off-gas treatment apparatus, in which the off-gas generated in the converter is recovered after the off-gas is cooled and dedusted, the state of the slag is determined based on the information detected by the acoustic instrument, the analysis of the off-gas composition, and dust densitometer, and control is performed based on the detection result, the height of the lance, the flow rate of the supplied oxygen, the ratio of the top and bottom purge, and the amount of auxiliary raw materials added to suppress the occurrence of a blowout and/or metal spatter.

List of cited documents

Patent Literature

PTL 1: Unexamined Japanese Patent Application No. 6-248321.

PTL 2: Unexamined Japanese Patent Application No. 59-41409.

PTL 3: Unexamined Japanese Patent Application No. 6-256832.

Disclosure of the essence of the invention

Technical problem

However, the above prior art technologies have the problems indicated below.

Patent Literature 1 uses a vibration sensor and Patent Literature 2 uses a microwave slag level meter. Thus, methods for predicting metal release using these sensors require that the sensors be installed inside the converter or in a position very close to the inside of the converter. Because the installed sensors are very close to the molten steel, slag, flue gas, etc., which move rapidly inside the converter during oxygen purge, there are problems with the durability and continuous operation of the equipment.

According to Patent Literature 3, measurement is performed in the exhaust gas path of the converter exhaust gas treatment equipment, and relatively stable operation can be realized because the temperature of the atmosphere in the exhaust gas path is low. However, a problem arises in that a delay occurs in the measurement time corresponding to the time during which the exhaust gas travels to the measurement point in the exhaust gas duct.

The present invention has been made in view of the prior art described above in order to achieve the following objects. One object of the present invention is to provide a method and system for predicting metal release during decarburization refining of molten pig iron in a converter without the need for a metal release sensor to be installed inside the converter or at a location very close to the inside of the converter, and without time delay. It is another object of the present invention to provide a method for operating a converter that allows oxygen to be blown through molten pig iron while preventing the occurrence of a metal blowout.

Solution

The essence of the present invention, designed to solve the problems described above, is as follows.

[1] A method for predicting a metal release in a converter occurring during decarburization refining in a converter in which molten pig iron is produced into molten steel by blowing an oxidizing gas into the molten pig iron in the converter from an upper blast lance, or blowing also with an oxidizing gas or an inert gas from a lower blast lance to perform decarburization refining of molten pig iron, characterized in that

measuring the radiation spectrum of the flame of the combustion process in the neck coming out of the neck of the converter;

calculate the radiation intensity of the measured radiation spectrum at a wavelength in the range of 580 - 620 nm; and

predicting the occurrence of a metal outburst based on a change, in a time series, of the calculated radiation intensity.

[2] The method for predicting metal ejection in a converter according to [1], further detecting an inflection point at which the radiation intensity starts to increase after it has previously decreased, and when said inflection point is detected, metal ejection is predicted.

[3] A converter metal emission prediction method according to [1] or [2], in which the change in radiation intensity in a time series is determined using a moving average.

[4] The converter metal emission prediction method according to [1] or [2], wherein the variation of the radiation intensity in the time series is determined according to a determination formula using a moving average.

[5] The converter metal emission prediction method according to [4], in which formulas (1) to (3) below are used as determination formulas, and when all formulas (1) to (3) are satisfied, the that the prediction result indicates the occurrence of a metal outburst.

[Mat. one]

where I(n, m ₀ ) is the moving average (arr. unit) of the radiation intensity index from the measurement point nm ₀ to the measurement point n, I(nL ₁ , m ₁ ) is the moving average (arr. unit) of the radiation intensity index from the point measurements nL ₁ -m ₁ to the measurement point nL ₁ , I(n-2L ₁ , m ₁ ) denotes the moving average (arbitrary unit) of the radiation intensity index from the measurement point n-2L ₁ -m ₁ to the measurement point n-2L ₁ , I(n, m ₂ ) denotes the moving average (arr. unit) of the radiation intensity index from the measurement point nm ₂ to the measurement point n, I(nL ₂ , m ₂ ) denotes the moving average (arr. unit) of the radiation intensity index from the point measurements nL ₂ -m ₂ to the measurement point nL ₂ , C ₀ , C ₁ and C ₂ are the thresholds of definition and satisfy the relations C ₀ > 0, C ₂ > 0 and C ₁ <C ₂ , L ₁ and L ₂ are constants, each of which is an integer, 1 or more, and m ₀ , m ₁ and m ₂ are constants, each of which is an integer chi scrapped, 0 or more.

[6] The converter metal emission prediction method according to [5], in which the detection threshold values of C ₀ , C ₁ and C ₂ in formulas (1) to (3) are determined using one or more of the following parameters: change in radiation intensity, off-gas flow rate, off-gas components, top blow lance oxygen feed rate, and top blow lance height during oxygen purge.

[7] The converter metal outburst prediction method according to [5], wherein the threshold values C ₀ , C ₁ and C ₂ in formulas (1) to (3) are determined by machine learning using one or more of the following parameters: intensity change emissions, off-gas flow rate, off-gas components, top lance oxygen supply rate, and top lance height during oxygen purge.

The operating method of a converter in which molten steel is obtained from molten pig iron, when a prediction result indicating the occurrence of a metal outburst is determined by a converter metal outburst prediction method according to any of [1] to [7], and after determining said prediction result, indicating the appearance of a metal ejection, at least one of the following adjustments is made: adjusting the flow rate of the oxidizing gas blown from the top blast lance, adjusting the height of the top blast lance, adjusting the position of the movable bell in height, adjusting the flow rate of oxidizing gas or inert gas blown from lower blow lance, and the addition of a soothing material.

[9] A system for predicting a metal release in a converter occurring during converter decarburization refining in which liquid steel is obtained from molten pig iron by blowing an oxidizing gas over the molten pig iron in the converter from an overhead blast lance or blowing also with an oxidizing gas or an inert gas from lower blow lance, to perform decarburization refining of molten pig iron, while the system contains:

a spectroscopic camera located near the converter with the possibility of obtaining an image of the flame of the combustion process in the neck through the gap between the converter and the movable bell; and

an image analyzer capable of recording image data, with the possibility of extracting them, sent from the spectroscopic camera, calculating from said image data the radiation intensity in the radiation spectrum at a wavelength in the range of 580 - 620 nm and predicting the metal emission based on the calculated change, in time row, the intensity of the radiation.

[10] The system for predicting the emission of metal in the converter according to [9], also comprising a control computer configured to output a control signal for changing the operating mode in accordance with data input from the image analyzer.

[11] A system for predicting metal ejection in a converter according to [9] or [10], in which the image analyzer is configured to detect an inflection point at which the emission intensity begins to increase again after it has previously decreased, and upon detection the specified inflection point to predict the specified occurrence of a metal outlier.

[12] A converter metal emission prediction system according to any one of [9] to [11], in which the image analyzer calculates the change in radiation intensity over time according to a moving average determination formula.

[13] A system for predicting a metal outburst in a converter according to [12], in which formulas (1) to (3) below are used as a determination formula, and a prediction result indicating the occurrence of a metal outburst is determined when all formulas (1) - (3).

[Mat. 2]

where I(n, m₀) denotes the moving average (arbitrary unit) of the radiation intensity index from the measurement point n-m₀ to the measuring point n, I(n-L_one, m_one) denotes the moving average (arbitrary unit) of the radiation intensity index from the measurement point n-L_one-m_one up to measuring point n-L_one, I(n-2L_one, m_one) denotes the moving average (arb. unit) of the radiation intensity index from the measurement point n-2L_one-m_oneto measuring point n-2L_one, I(n, m₂) denotes the moving average (arbitrary unit) of the radiation intensity index from the measurement point n-m₂ to the measuring point n, I(n-L₂, m₂) denotes the moving average (arb. unit) of the radiation intensity index from the measurement point n-L₂-m₂ up to measuring point n-L₂, C₀, C_one and C₂ are thresholds of definition and satisfy the relations C₀> 0, C₂> 0 and C_one<C₂, L_one and L₂ are constants, each of which is an integer, 1 or more, and m₀, m_one and m₂ are constants, each of which is an integer, 0 or more.

[14] A converter metal ejection prediction system according to [13], in which the image analyzer includes a machine learning model to determine the threshold values of C ₀ , C ₁ and C ₂ in formulas (1) to (3) by machine learning using one or several of the following: changing the radiation intensity, the off-gas flow rate, the off-gas components, the oxygen gas feed rate from the top blow lance, and the height of the top blow lance during oxygen purge.

[15] The system for predicting the release of metal from the converter according to [13], also containing a machine learning computer, which includes a machine learning model for determining threshold values for determining C ₀ , C ₁ and C ₂ in formulas (1) - (3 ) by machine learning using at least one of the following parameters: change in radiation intensity, off-gas flow rate, off-gas components, oxygen gas supply rate from the top blow lance, and height of the top blow lance during oxygen purge.

Useful effects of the invention

In the method and system for predicting metal ejection in a converter according to the present invention, the possibility of metal ejection is predicted by measuring the emission spectrum of the throat flame. Therefore, the appearance of a metal surge can be predicted without the need for a metal surge detector to be installed inside the converter or at a position very close to the inside of the converter, and without creating a time delay. In the operation method of the converter according to the present invention, since metal ejection prevention measures are performed at the time when a prediction result indicating the occurrence of a metal ejection event is obtained, the metal ejection event can be reliably suppressed.

Brief description of the drawings

Fig. 1 is a diagram illustrating the hardware configuration of a converter suitable for implementing the present invention.

Fig. 2 is a graph illustrating the change in time series of the radiation intensity index during an oxygen purge in a melt in which a burst occurred.

Fig. 3 is a graph illustrating the change in the time series of the radiation intensity index during oxygen purge in a melt in which no emission occurred.

Fig. 4. is a diagram illustrating the hardware configuration of a converter suitable for implementing the present invention

Description of embodiments of the invention

The inventors have extensively researched to predict, in the decarburization refining performed in a converter for producing liquid steel from molten pig iron by oxidative refining of molten iron, in real time without time delay, the occurrence of a metal burst during oxygen purging. In particular, during the decarburization refining in the converter, the inventors monitored the internal state of the converter when metal ejection occurred in real time. It is known that a blowout occurs when the slag in the converter is in a foamed state.

As a result of research, the inventors focused on the flame of the combustion reaction in the neck of the converter as a factor suitable for accurately determining the internal state of the converter in real time, and came up with the idea of measuring the emission spectrum of the flame of the combustion reaction in the neck at predetermined time intervals during decarburization refining. Here, the term "throat combustion reaction flame" means the flame inside the converter, the flame exiting the neck of the converter towards the chimney on top of the converter.

The emission spectrum of the combustion reaction flame in the throat contains not only information about the gaseous CO formed as a result of the decarburization reaction (C + O → CO) in the converter, but also information about the gaseous CO ₂ formed as a result of spontaneous combustion resulting from the mixing of part of the formed gaseous CO and air sucked in through the neck of the converter. The emission spectrum additionally contains information on FeO* (intermediate product) related to the iron atoms that evaporate from the ignition point inside the converter (namely, from the position at which the oxidizing gas from the top blow lance hits the surface of the molten pig iron bath). The inventors have found that the internal state of the converter can be easily assessed in real time if it is possible for a wavelength in the range of 580 to 620 nm in the emission spectrum to measure the emission intensity in real time at each wavelength.

The wavelength in the range of 580 - 620 nm in the emission spectrum corresponds to the orange band of the FeO spectrum, associated with the formation and disappearance of FeO * (intermediate product), and differs from the wavelengths of the band of the intermediate gas product containing gaseous hydrocarbons. In addition, the inventors confirmed that an absorption peak is observed in the above wavelength range when FeO* (intermediate) is formed, while an emission peak is observed in the same wavelength range when FeO* (intermediate) disappears. In addition, the inventors have confirmed that the intensity of the radiation correlates with the rate of disappearance of FeO* (intermediate). Hereinafter, "FeO* (intermediate)" is simply referred to as "FeO*".

From the above point of view, the inventors measured the emission spectrum of the combustion reaction flame in the neck of the converter in a time series during the decarburization refining in the converter. The emission spectrum of the combustion reaction flame in the converter was measured as shown in FIG. bell 10. The image acquired by the spectroscopic camera 6 is transmitted to the image analyzer 7. In the image analyzer 7, the image is recorded and the emission intensity at each emission wavelength is analyzed by performing a linear analysis of the input image data on any scan line. The emission spectrum was measured and the intensity of the emission was analyzed with an interval between measurement points, that is, with a measurement time interval Δt taken equal to 1 second.

Based on the obtained results of measuring the emission spectrum, the wavelength of 610 nm, at which the width of the change was largest during decarburization refining, was selected as the target wavelength (the wavelength used in the analysis), and the change in the emission intensity in the time series was obtained by calculating radiation intensity at a wavelength of 610 nm at each measurement moment during decarburization refining. When obtaining the change in the emission intensity in the time series, the normalized emission intensity was determined as the "radiation intensity index", assuming that the emission intensity in the image data recorded by the spectroscopic camera 6 in the throat before the start of the oxygen purge is determined as 1, and the emission intensity change in the time series was obtained using this radiation intensity index. Of course, the change in the time series can also be obtained using non-normalized radiation intensity.

In the studies described above, a converter (capacity of 300 tons) was used, allowing the supply of purge oxidizing gas from the upper blow lance 3 and the supply of purge agitation gas from the bottom blow lance 4 at the bottom of the converter. Oxygen gas (commercial grade oxygen gas) was used as the oxidizing gas blown from the top blow lance, and argon gas was used as the stirring gas blown from the bottom blow lance. The top blow lance used here was a top blow lance including Laval blow nozzles serving as five blow nozzles for oxygen gas, mounted at the end of the lance with a blow angle of 15º. Here, the blowing angle for the blow nozzle represents the relative angle between the blowing direction of oxygen gas from the blow nozzle and the axial direction of the upper blow lance.

In the converter described above, the decarburization refining was carried out on liquid pig iron with a carbon concentration of 3.5% by weight. The supply of gaseous oxygen from the upper blast lance began at the moment when the carbon content in the molten pig iron was 3.5 wt%. and continued until the carbon content in the molten iron in the converter did not reach 0.04% of the mass.

The supply rate of oxygen gas from the upper blow lance was set to 800 to 1000 Nm ³ (Nm)/min, the height of the lance of the top blow lance was set in the range of 2.5 to 3.0 m, and the agitation gas supply rate from the bottom blow lance was set to equal to 5-30 nm ³ /min. In the description, the term "upper blast lance height" means the distance from the head of the upper blast lance to the surface of the molten pig iron bath inside the converter when measured at steady state.

Fig. 2 is a graph illustrating the change in the time series of the radiation intensity index during oxygen purge during the melt in which the metal was ejected, the radiation intensity index was calculated in accordance with the method described above. Fig. 3 is a graph illustrating the change in the time series of the radiation intensity index during oxygen purge during a melt in which no metal was ejected, with the radiation intensity index calculated in accordance with the method described above. The decarburization refining time during the melt illustrated in FIG. 2 in which the burst occurred was 19.5 minutes, and the decarburization refining time during the melt illustrated in FIG. 3, in which no ejection occurred, was 18.0 min. The oxygen purge stroke, represented by the horizontal axis in each of FIGS. 2 and 3 is determined by the following formula (4).

Oxygen purge stroke = (Q _O2C / Q _0O2 ) × 100… (4)

where Q _O2C denotes the total amount of oxygen (nm ³ ) during the period of time from the start of the oxygen supply to a certain point in time, and Q _O2 denotes the total amount of oxygen (nm ³ ) at the end of the oxygen supply.

As can be seen from FIG. 2 and 3, as the oxygen purge continues, the radiation intensity index increases in the first half of the oxygen purge (namely, the period of time when the progress of the oxygen purge is in the range of 60-70%), regardless of the occurrence of a burst. On the other hand, in the second half of the oxygen purge, as the oxygen purge is performed, the radiation intensity index decreases.

However, as illustrated in FIG. 2, in the melt in which the metal ejection occurred, the radiation intensity index, which starts to increase during the oxygen purge, sometime decreases even in the first half of the oxygen purge, and then ejection occurs after the emission intensity index starts to increase again.

The above phenomenon is considered to be the result of the fact that at the time of occurrence of the metal ejection, since the slag in the converter causes foaming, the apparent thickness of the slag increases and the FeO reduction reaction slows down, i.e. the decarburization reaction is slowed down due to the blocking effect caused by the increase in the apparent thickness of the slag, as a result of which the radiation intensity index is reduced once. The reason why the radiation intensity index increases again after that is that since the FeO reduction reaction has stopped, the amount of FeO in the slag becomes excessive and the decarburization reaction (FeO + C→Fe + CO) starts again at the interface between the slag and molten iron, thereby again increasing the radiation intensity index.

From the above result, the inventors have found that a change in the time series of the radiation intensity index can be used to predict the emission of metal.

During a melt that did not eject metal, as shown in FIG. 3, the radiation intensity indicator increases as the purge is carried out and reaches its maximum value at the middle stage of the purge. Then, in the period approaching the final purge stage, the emission intensity index decreases as the rate of the iron oxide reduction reaction decreases.

In view of the fact that the radiation intensity index forms an ascending and descending pattern, the inventors compared the change in the time series of the radiation intensity index between the melt in which the emission occurred and the melt in which the emission did not occur. As a result, it was found that in the case of a metal ejection, a change in the time series of the radiation intensity index has, for example, the feature mentioned below. Metal ejection occurs when the current value of the radiation intensity index increases by 20% or more from the value of the radiation intensity index at the measurement point 10 seconds before the current time, and the radiation intensity index at the measurement point 10 seconds before the current time is equal to or less than in measuring point 80 seconds before the current time. The above consideration applies similarly to measured radiation intensity values that are non-normalized.

Thus, it has been suggested that when an ejection occurs, the following pattern of change in radiation intensity and radiation intensity index is observed: each of them continuously decreases or does not change for a certain time (about 70 seconds), and then proceeds to a significant increase, as described above (the nature of the change is defined as the "inflection point" in this description). In other words, it has been suggested that the occurrence of a metal ejection can be predicted by detecting the appearance of an inflection point in the time series of changes in radiation intensity and radiation intensity index.

In the melt shown in Fig. 2, in which metal ejection occurs, there are two periods (radiation intensity index troughs) in which the emission intensity index decreases once and then begins to increase. The rate of increase in the radiation intensity index for each of the two pits compared to the radiance intensity index at the measurement point 10 seconds before the measurement point corresponding to each pit is slower in the first pit (at the time when the oxygen purge is approximately 30% complete) than in the second trough (when the oxygen purge is approximately 45% complete). Thus, the first trough (at the time when the oxygen purge is approximately 30% complete) is considered to be the expected occurrence of slag foaming that does not result in a blowout. On the other hand, in the second valley (at the time when the oxygen purge is approximately 45% completed), the increase rate of the radiation intensity index compared to the radiation intensity index at the measurement point 10 seconds before exceeds 20%, and the metal ejection occurs after , as the radiation intensity index passed the second trough. Therefore, it is believed that the occurrence of a metal ejection can be more accurately predicted by detecting the occurrence of an inflection point, such as a second trough, in the time series of changes in radiation intensity and radiation intensity index.

When using the radiation intensity index to predict metal ejection, metal ejection can also be predicted by comparing the radiation intensity index given as an instantaneous value (actual value) that is not a running average, as seen from the mode illustrated in FIG. 2 and 3. However, it has been confirmed that metal release can be more accurately predicted using a moving average of the radiation intensity index over a certain period. Here, the term "moving average" means the value obtained by dividing the sum of data changes over a certain period of time by the number of data units, and is a method of smoothing time series data.

The deviation is reduced by using moving averages of the radiation intensity (actual values) and the radiation intensity index. With an appropriate choice of the number of readings in the calculation of the moving averages, in a melt in which there is no ejection of metal, the moving average can be caused, for example, by a monotonic increase until the radiation intensity index reaches its maximum value, and a monotonic decrease after the radiation intensity index has reached its maximum value. The radiation intensity (actual value) also behaves similarly to the radiation intensity index.

In addition, it has been confirmed that the occurrence of a metal outburst can be predicted even more accurately by calculating the changes in the time series of the radiation intensity and the radiation intensity index according to the moving average determination formulas.

For example, the following formulas (1) to (3) can be used as moving average determination formulas of the radiation intensity index with respect to the measurement point n. Using formulas (1) - (3), you can easily find the above inflection point. Here, the measurement point n denotes a measurement point at a certain time during decarburization refining, and corresponds to the current measurement point.

[Mat. 3]

In the above formulas, I(n, m ₀ ) denotes the moving average (arr. unit) of the radiation intensity index from the measurement point nm ₀ to the measurement point n, I(nL ₁ , m ₁ ) denotes the moving average (arr. unit) of the intensity index radiation from the measurement point nL ₁ -m ₁ to the measurement point nL ₁ ; I(n-2L ₁ , m ₁ ) denotes the moving average (arr. unit) of the radiation intensity index from the measurement point n-2L ₁ -m ₁ to the measurement point n-2L ₁ ; I(n, m ₂ ) denotes the moving average (arr. unit) of the radiation intensity index from the measurement point nm ₂ to the measurement point n; I(nL ₂ , m ₂ ) denotes the moving average (arr. unit) of the radiation intensity index from the measurement point nL ₂ -m ₂ to the measurement point nL ₂ ; C ₀ , C ₁ and C ₂ are detection thresholds and satisfy the relationships C ₀ >0, C ₂ > 0 and C ₁ <C ₂ ; L ₁ and L ₂ are constants, each of which is an integer, 1 or more, and m ₀ , m ₁ and m ₂ are constants, each of which is an integer, 0 or more.

Here, the formula (1) implies that only data with a C value of ₀ or more is used as determination data. With such a condition, the determination can be made by excluding background noise and data measured during the blackout period when the field of view of the spectroscopic camera was blocked for a short time. Formula (2) represents the change in the radiation intensity index during the period of time from the measurement point n-2L ₁ to the measurement point nL ₁ , i.e. over a period of time shortly before the current point in time. Formula (3) represents the magnitude of the change in the radiation intensity index during the period of time from the measurement point nL ₂ to the measurement point n, i.e. over a period of time shortly before the current point in time.

Formulas (2) and (3) are normalized using I(n-2L ₁ , m ₁ ) and I(nL ₂ , m ₂ ) respectively. The normalization is intended to eliminate the effects caused by the absolute value of the radiation intensity, which varies for each heat. In addition, in the present invention, due to the detection of an inflection point in the change in the time series, i.e. phenomena when the radiation intensity index decreases once in a time period, shortly before the current time, and then returns to increase just before the current time, C ₂ > 0 and C ₁ <C ₂ are performed.

Each of L ₁ and L ₂ represents the number of measurement points since the current time. Assuming that the measurement time interval is Δt (sec), each of L ₁ ×Δt and L ₂ ×Δt represents a period (in seconds) counted from the current time. Moreover, each of m ₀ , m ₁ and m ₂ represents the number of measurement points falling within the range of the inverse moving average. Assuming that the measurement time interval is Δt (sec), m ₁ ×Δt, m ₂ ×Δt and m ₃ ×Δt, each represents the time interval (in seconds) over which the inverse moving average must be calculated.

Assuming that metal ejection occurs when all formulas (1) to (3) are satisfied, tests for predicting the occurrence of metal ejection during decarburization refining according to formulas (1), (2) and (3) above were carried out by varying threshold values for determining C ₀ , C ₁ and C ₂ and constants L ₁ , L ₂ and m ₀ , m ₁ and m ₂ .

Table 1 presents the test results. In these tests, a wavelength of 610 nm was set as the target wavelength, and even when a prediction result indicating the occurrence of metal ejection was determined, the operation mode was not changed to prevent the occurrence of metal ejection.

Table 1

Constant Threshold Definition result L1 _,
L2 m ₀ ,
m ₁ ,
_{m2 C0 C1 C2} Smelting with ejection Normal melting Percentage of successful determinations (%) Percentage of normal detection
(%) Level 1 ten twenty fifteen 0.65 0.7 68 82 Level 2 ten ten fifteen 0 1.0 52 65 Level 3 fifteen fifteen ten 0 0.8 45 62 Level 4 fifteen twenty fifteen 0.65 0.7 53 89 Level 5 ten fifteen ten 0.65 0.7 62 72 Level 6 ten fifteen 5 0.65 0.7 64 42

Number of heats performed: 100 heats for each test.

As can be seen from Table 1, it has been found that metal ejection can be stably predicted by appropriately selecting the number of readings in the moving average calculation of the radiation intensity index and detection thresholds. Here, "Determination Success Rate" in Table 1 refers to the percentage of heats for which a blowout was predicted 60 seconds or more before the blowout actually occurred. "Normal Detection Fraction" means the percentage of heats for which a prediction result indicative of the occurrence of a metal outburst was not determined for heats that did not have a metal outburst, i.e. percentage of heats that did not have a false positive.

A converter metal emission prediction method, a converter metal emission prediction system, and a converter operation method according to the present invention have been developed by further studies based on the above-described data. Details of the execution of the converter metal emission prediction method, the converter metal emission prediction system, and the operation method of the converter according to the present invention will be described below with reference to the drawings. 1 is a schematic view illustrating the hardware configuration of a converter suitable for carrying out the present invention.

Converter 1 equipment suitable for carrying out the present invention includes: converter 2; upper blow lance 3; a spectroscopic camera 6 located on the periphery of the converter 2 to be able to capture the image of the flame 16 of the combustion reaction in the neck; a recording image analyzer 7 capable of obtaining an image captured by the spectroscopic camera 6 and analyzing the captured image; and a control computer 8 sending control signals based on the data analyzed by the image analyzer 7.

The equipment of the converter 1 further comprises a lance height controller 11 for adjusting the height of the upper blow lance 3 and an oxidizing gas flow controller 12 for controlling the feed rate of the oxidizing gas which is blown from the top blow lance 3, these controllers operating individually in accordance with control signals sent from control computer 8. Converter 1 equipment additionally includes a lower purge gas flow controller 13 to control the mixing gas supply rate blown from the lower blow tuyere 4, a auxiliary raw material addition controller 14 to control the grades and amount of auxiliary raw materials added, which are stored in bins (not shown ) located above the converter, and the regulator 15 of the position of the movable bell to control the height position of the movable bell 10.

The lance height measured by the lance height controller 11, the oxidizing gas feed rate measured by the oxidizing gas feed rate controller 12, the bottom blowing gas feed rate measured by the bottom purge gas rate controller 13, the amount of added auxiliary raw material measured by the auxiliary raw material addition controller 14, and the height position of the movable bell measured by the movable bell position controller 15 are input to the control computer 8 for feedback control. In addition, the flue gas flow rate measured by a flue gas flow meter (not shown) installed in the chimney and measuring the flue gas flow rate leaving the converter and the content of flue gas components measured by a flue gas analyzer (not shown) located in the flue, and the determined content of the components (CO, CO ₂ and O ₂ ) of the off-gas leaving the converter are additionally entered into the control computer 8.

In the converter 2 used in the present invention, the blown oxidizing gas stream 17 can be directed from the top blow lance 3 towards the molten pig iron 5 in the converter, and at the same time the stirring gas can be blown into the converter from the bottom blow lance 4 at the bottom. converter. In addition, a spectroscopic camera 6 is installed near the converter, capable of measuring the emission spectrum of the flame 16 of the combustion reaction in the neck in the converter 2. The image of the flame 16 of the combustion reaction in the neck, appearing through the gap between the neck 9 of the converter and the movable bell 10 is recorded by the installed spectroscopic camera 6.

The installation position of the spectroscopic camera 6 can be further chosen so that the impact of heat, dust, etc., on the spectroscopic camera 6 is low enough to make the spectroscopic camera 6 durable and the spectroscopic camera 6 can capture the image of the combustion reaction flame 16 in the throat, emerging through the gap between the neck 9 of the converter and the movable bell 10. For example, when the spectroscopic camera 6 is installed in front of the converter 2, the image of the flame 16 of the combustion reaction, appearing through a small window (gap) located in the loading hatch, to check the flame, can be recorded by spectroscopic camera 6. Instead, the image registration window through which the image of the flame 16 of the combustion reaction in the throat can be obtained can be located on the rear side (opposite the control side) of the wall surrounding the converter body, or on the side of the converter (from the side of the support roller) , and a spectroscopic camera 6 can be placed outside the window to capture an image of the combustion reaction flame 16 in the throat. Alternatively, the spectroscopic camera 6 may be installed even inside the wall surrounding the converter housing, if the spectroscopic camera 6 is capable of withstanding the conditions of this position.

The image (image data) captured by the spectroscopic camera 6 is sequentially sent to the image analyzer 7. The image analyzer 7 records the sent captured image (image data), performs linear analysis of the image data in any scan line, and analyzes the emission wavelengths and emission intensity for each wavelength. radiation waves. In addition, the image analyzer 7 predicts the occurrence of a metal ejection based on the analysis result of the sent captured image (image data).

The image data of the throat combustion reaction flame 16 analyzed by the image analyzer 7 is sent to the control computer 8 at each measurement moment along with a prediction result indicating the occurrence of a metal ejection. Similarly, operating data stored in the control computer 8, such as the top lance oxygen gas feed rate, top lance height, off-gas flow rate, and off-gas components, are sent to the image analyzer 7 with each measurement.

After receiving from the analyzer 7 an image of the prediction result indicating the occurrence of a metal ejection, the control computer 8 outputs control signals for the operation of the lance height controller 11, the oxidizing gas flow controller 12, the gas flow controller 13 blown from below, the auxiliary raw material supply controller 14 and the controller 15 position of the movable bell, independently or jointly. Position 18 in FIG. 1 denotes a pipe for supplying oxidizing gas to the top blast lance, numeral 19 denotes a pipe for supplying cooling water to the top blast lance, and numeral 20 denotes a pipe for draining cooling water from the top blast lance.

According to the present invention, liquid steel is produced from molten pig iron 5 by means of converter 1 equipment by blowing an oxidizing gas into the converter from an upper blast lance 3 or additionally blowing also inert gas from a lower blast lance 4 through the molten pig iron 5 contained in the converter 2, and by carrying out the oxidative refining of the molten pig iron 5, i. e. decarburizing refining of molten pig iron 5.

During the decarburization refining, an image of the combustion reaction flame 16 in the throat is captured by the spectroscopic camera 6, and the resulting emission spectrum is analyzed to evaluate the change in the situation inside the converter 2 during the real-time decarburization refining. The release of metal is predicted based on the expected change in the situation inside the converter. From the point of view of increasing productivity and increasing the yield of iron, it is preferable to record the image of the combustion reaction flame 16 in the neck with the spectroscopic camera 6 and analyze the emission spectrum in the measurement time interval Δt, set equal to 1 - 10 s.

The emission spectrum obtained by capturing an image of a flame is recorded in the image analyzer 7 in an extractable manner. The image analyzer 7 determines the emission wavelengths in the range of 580-620 nm in the obtained emission spectrum of the throat combustion reaction flame 16 and calculates the emission intensity for each of the specified emission wavelengths.

As described above, the wavelength in the range of 580 - 620 nm corresponds to the orange band of the FeO system, which is associated with the formation and disappearance of FeO*. The inventors confirmed that an absorption peak is observed when FeO* is formed in the above wavelength range, while an emission peak is observed in the same wavelength range when FeO* disappears, and the emission intensity correlates with the FeO* disappearance rate. In other words, since wavelengths in the range of 580 - 620 nm reflect the reaction in the converter and provide information that makes it easy to assess the situation inside the converter, these wavelengths are chosen as the measurement target. The radiation intensity indicates the amount of radiation energy when FeO goes from the excited state (FeO*) to the ground state.

The image analyzer 7 calculates the radiation intensity and the resulting radiation intensity index for each wavelength. In addition, the image analyzer 7 preferably calculates a moving average of the radiation intensity index. Then, the image analyzer 7 evaluates the change in the situation inside the converter and predicts the occurrence of a metal ejection during the operation of the converter based on the calculated changes in the time series of the radiation intensity, the radiation intensity index, and the moving average of the radiation intensity index. In this case, the above-described inflection point is preferably detected, and the occurrence of a metal ejection is predicted according to the detection result indicating the presence of the inflection point.

Preferably, formulas (1) to (3) described above are used as determination formulas for predicting moving average metal outburst, and a prediction result indicating the occurrence of metal outburst is determined when all formulas (1) to (3) are executed. Since C ₀ , C ₁ and C ₂ , i.e. detection thresholds in formulas (1) to (3) differ depending on the image capture conditions and the operating mode of each converter, a preliminary test is performed to measure the radiation intensity and to set the detection thresholds so that the percentage of normal detection of a metal emission based on formulas (1) - (3) was the maximum. Considering the performance advantage, C ₀ - C ₂ can also be defined such that the normal detection rate is maximum within the range in which false detection is kept the least.

As for the other constants L ₁ , L ₂ , m ₀ , m ₁ and m ₂ , when each constant is set to a larger value, the detection trend becomes more moderate and the over-detection (the case of determining a non-ejection state as causing metal ejection) is less probable. However, if the values of these constants are set too high, the sensitivity decreases and it becomes difficult to detect a metal outlier before it occurs.

In addition, ∆t × L ₁ , ∆t × L ₂ , ∆t × m ₀ , ∆t × m ₁ , and ∆t × m ₂ are preset to obtain the maximum percentage at which normal metal emission detection is performed according to formulas (1) - (3) by performing a preliminary test taking into account the time (from several tens of seconds to one hundred seconds) required to take countermeasures after detecting the occurrence of an outlier. When each of Δt×L ₁ , Δt×L ₂ , Δt×m ₀ , Δt×m ₁ , and Δt×m ₂ is set to a duration corresponding to about 1 to 5% of the purge time, a fairly suitable degree of detection is obtained. Considering the performance advantage, these constants can also be defined such that the proportion of normal detection is maximized within the range in which false detection is kept the least.

Using the image analyzer 7, the operator can set the threshold values C ₀ , C ₁ and C ₂ and the constants L ₁ , L ₂ , m ₀ , m ₁ and m ₂ equal to any numerical values. In addition, the image analyzer 7 has a machine learning model function for determining threshold values of C ₀ , C ₁ and C ₂ by machine learning using one or two or more changes in radiation intensity, exhaust gas flow rate, exhaust gas components, oxygen gas supply rate into the top lance and heights of the top lance during oxygen purge. In other words, the image analyzer 7 has a function to enable automatic adjustment of C ₀ , C ₁ and C ₂ through machine learning.

Further, as illustrated in FIG. 4, the machine learning computer 21 with the machine learning model function can be placed separately from the image analyzer 7. In this case, the above-described constants can be set as follows. First, for example, from the control computer 8 in which the operation data is recorded and from the image analyzer 7 in which the emission intensity data of the throat combustion reaction flame 16 is recorded, respectively, the operation data and the emission intensity data are sent to the machine learning computer. 21 offline. The machine learning computer 21 performs machine learning based on the received data, determines new values of the above constants, and sends the determined values to the image analyzer 7 . The image analyzer 7 receives the new constants and makes a determination in the next and subsequent operations using the new constants.

4 is a schematic view illustrating another hardware configuration of a converter suitable for implementing the present invention. Converter equipment 1A shown in FIG. 4 has a configuration in which the machine learning computer 21 is located in addition to the converter equipment 1 shown in FIG. 1. Another configuration is similar to the hardware configuration of the converter 1 shown in FIG. 1. The same components are designated by the same reference numerals, and the description of these components is omitted.

The specific wavelength used in calculating the radiation intensity index is determined by preliminary measurement of wavelengths in the range of 580 - 620 nm at the one wavelength at which the amount of change in radiation intensity during decarburization refining is the largest, or by periodically monitoring wavelengths within the above wavelength range during decarburization refining and determining for each monitoring one of several wavelengths at which the amount of change in radiation intensity is greatest.

In the operation method of the converter according to the present invention, at the time when the prediction result indicating the occurrence of a metal outburst is determined during decarburization refining based on the calculated change in the radiation intensity time series, one or two or more actions of adjusting the supply rate of the oxidizing agent of gas blown from the top blow lance, adjusting the height of the lance, adjusting the height position of the movable bell, adjusting the feed rate of the oxidizing gas or inert gas blown from the bottom blow lance, and adding a soothing material to the converter is performed to prevent blowout and equipment damage. In this case, the change in the time series of the radiation intensity index is preferably calculated using the determination formulas (1) to (3), and the prediction result indicating the occurrence of a metal outburst is determined when all formulas (1) to (3) are satisfied.

The practical method of combating the occurrence of metal ejection is preferably carried out by reducing the supply rate of the oxidizing gas blown from the top blowing lance, reducing the height of the top blowing lance, increasing the height position of the movable bell to prevent damage to the movable bell by slag, increasing the supply rate of the oxidizing gas or inert gas blown from the lower blow lance, or the addition of a soothing material. More preferably, two or more of the operations described above are performed concurrently. With the adjustment described above, foaming of the slag or a strong decarburization reaction can be suppressed before the ejection is possible or at the very beginning of the ejection of the metal. Therefore, the discharge of slag and liquid iron outside the converter can be avoided, and the yield of iron can be improved.

In this case, decreasing the supply rate of the oxidizing gas injected from the top blow lance, decreasing the height of the top blow lance, increasing the supply rate of the gas blown from the bottom blow lance, and adding a soothing material are countermeasures to change the operating conditions to prevent metal blowout. Raising the height of the movable bell is a measure against changing working conditions to prevent damage to the equipment. Thus, it is desirable to change at least the operating conditions in order to prevent the occurrence of metal ejection. Reducing the height of the movable bell to physically prevent the release of slag is also an effective countermeasure in terms of reducing the release of slag. In such a case, however, the reduction in the number of times that the movable bell can be used must be taken into account.

Here, the term "calming material" means a steelmaking auxiliary raw material that suppresses slag foaming by forming degassing paths in the slag inside the converter when added to the converter and by improving slag degassing when foaming. Usually, for example, carbonaceous material, mill scale or slag is used in granular form after mixing with moisture or oil and fat, but any other suitable substance can also be used.

Preferably, the reduced rate of supply of oxidizing gas blown from the top blow lance, the reduced amount of height of the top blow lance, the amount of height change of the movable bell, the increased rate of supply of oxidizing gas or inert gas blown from the bottom blow lance, and the amount of soothing material added are determined by in advance, for example, based on the relationship between the mixing intensity of the molten iron and the flow rate of the oxidizing gas.

In the converter equipment 1 suitable for realizing the present invention, preferably every time all determination formulas (1) to (3) are executed, the control computer 8 outputs a control signal to the lance height controller 11 to decrease the lance height, outputs a control signal to the controller 12 the flow rate of oxidizing gas to reduce the flow rate of oxidizing gas blown from the upper blast lance, gives a control signal to the regulator 15 of the position of the movable bell in height to increase the height position of the movable bell, gives a control signal to the regulator 13 of the gas flow rate blown from below to increase the flow rate of the blown oxidizing gas or inert gas, outputs a control signal to the auxiliary raw material addition controller 14 to add a predetermined amount of soothing material, or all of these control signals at the same time.

The oxidizing gas blown from the top blowing lance 3 is usually oxygen gas (industrial grade oxygen gas). However, for example, a gas mixture of oxygen gas, an inert gas such as argon or helium, and nitrogen gas, air or oxygen-enriched air can also be used instead. The term "oxidizing gas" means an oxygen-containing gas with an oxygen concentration greater than or equal to the oxygen concentration in air. The gas injected from the lower blow lance 4 is an inert gas or an oxidizing gas. When an oxidizing gas is introduced, the oxidizing gas acts not only as an oxidizing gas for oxidizing refining, but also as an agitating gas.

When determining the situation inside the converter based on the spectral analysis of the combustion reaction flame 16 in the neck, false detection may occur in some cases due to a change in the situation, for example, in the case where the field of view is blocked by the passage of a crane. or metal deposits in the throat. Therefore, the threshold values for determining C ₀ , C ₁ and C ₂ in formulas (1) - (3) it is desirable to change during the operation of the converter for each heat.

In more detail, the C ₀ , C ₁ and C ₂ detection thresholds are preferably determined using one, or two, or more changes in irradiance, off-gas flow rate, off-gas components, top lance oxygen gas feed rate, and top lance height. oxygen purge time.

More preferably, C ₀ , C ₁ , and C ₂ detection thresholds are determined by machine learning using one or two or more changes in irradiance, off-gas flow rate, off-gas components, top lance oxygen gas feed rate, and top tuyere height. blowing lance during oxygen purge.

Although in the above description, the calculations according to formulas (1) to (3) are performed using the radiation intensity index, the calculations according to formulas (1) to (3) can be performed using the radiation intensity at each measurement.

According to the present invention, as described above, since the occurrence of a metal emission in the converter 2 performing decarburization refining of pig iron 5 is predicted by measuring the emission spectrum of the throat combustion reaction flame, it is no longer necessary to place a metal emission detection sensor inside the converter or at a location , very close to the inside of the converter, and the occurrence of metal ejection can be predicted without time delay. In addition, since countermeasures for preventing metal ejection are performed at the same time that a prediction result indicating the possibility of metal ejection is obtained, the occurrence of metal ejection can be stably suppressed.

EXAMPLE 1

The decarburization refining of molten pig iron 5 is carried out using a top and bottom purge converter (oxygen gas blown from above and argon gas blown from below) with a capacity of 300 tons and a configuration similar to that of converter 2 shown in FIG. 1. The top blow lance 3 is represented as a top blow lance comprising five Laval nozzles mounted at the end of the lance with a blowing angle of 15º, the nozzles being equally spaced on the same circle around the axis of the top blow lance. Each of the blow nozzles has a throat diameter d _t of 73.6 mm and an outlet diameter d _e of 78.0 mm.

First, iron scrap is loaded into the converter, and then molten pig iron weighing 300 tons, previously subjected to desulfurization and dephosphorization, is loaded into the converter at a temperature of 1310 - 1360ºС. Table 2 shows the chemical composition of molten pig iron.

table 2

Chemical composition of molten pig iron
(% mass) Molten pig iron temperature (°C) C Si Mn P S Cr Fe 3.4-
3.6 0.01 -
0.02 0.16 -
0.27 0.015 -
0.036 0.008 -
0.016 traces the rest
noe 1310 - 1360

Then, refining of the molten pig iron by decarburization is started by blowing oxygen gas as an oxidizing gas to the surface of the molten pig iron bath from the upper blowing lance 3, while argon gas is blown as a stirring gas into the molten pig iron bath from the lower blowing lance 4. Quantity loaded iron scrap is regulated so that the temperature of the molten steel after the end of decarburization refining was maintained at 1650°C.

Thereafter, quicklime is added as a CaO-based flux during the decarburization refining from a hopper (not shown) located above the converter, and the decarburization refining is continued until the carbon concentration in the molten iron reaches 0.05 wt%. The amount of added quicklime is adjusted so that the basicity ((CaO % wt.)/(SiO ₂ % wt.) of the slag formed in the converter is maintained at 2.5.

During decarburization refining, the image of the flame 16 of the combustion reaction in the neck, which occurs in the gap between the neck 9 of the converter 2 and the movable cap 10, is successively recorded with a predetermined measurement time interval Δt: 1 s using a spectroscopic camera 6 located practically in front of the converter 2.

The image analyzer 7 measures the emission spectrum (image data) from the captured image, and analyzes the predetermined emission wavelengths in the range of 580 to 620 nm in the acquired emission spectrum at each measurement time, and calculates an emission intensity index for each wavelength. The wavelength used here (ie, the target wavelength) is 610 nm. The analysis is performed by linear analysis of image data in any scan line.

Calculations in accordance with the above-described formulas (1) - (3) are performed using the radiation intensity index obtained at a specific wavelength at each measurement moment. In this case, the threshold values in the definition formulas (1) - (3) are set to C ₀ = 15, C ₁ = 0.65, and C ₂ = 0.7, and the constants are set to L ₁ = L ₂ = 10 and m ₀ = m ₁ \u003d m ₂ \u003d 20.

A prediction result indicating the occurrence of a metal outburst is determined when all formulas (1) to (3) are executed. When a prediction result is determined indicating the occurrence of a metal outburst, then one, two or more of the following are performed: adjusting the supply rate of gaseous oxygen from the upper blow lance, adjusting the height of the lance of the upper blow lance, adjusting the height position of the movable bell, adjusting the supply rate of the oxidizing gas or inert gas injected from the lower blow lance and adding a soothing material to the converter.

More specifically, the supply rate of oxygen gas from the top blowing lance is controlled by decreasing from 1000 Nm ³ /min to 833 Nm ³ /min, the height of the lance is adjusted by decreasing from 3.0 m to 2.5 m, and the supply rate of gas blown from below is adjusted , increasing from 15 nm ³ /min to 30 nm ³ /min. The height position of the movable bell is adjusted by raising the position of the movable bell by 500 mm at the time when a prediction result indicating the occurrence of a metal ejection is determined. The amount of added soothing material is adjusted to 500 - 1500 kg.

The above-described adjustments are made by supplying control signals from the control computer 8 to the lance height controller 11, the oxidizing gas flow controller 12, the bottom injection gas flow controller 13, the auxiliary raw material addition controller 14 and the movable bell position controller 15 to control these controllers immediately after being the prediction result is determined, indicating the occurrence of a metal outburst. In addition, when any of the formulas (1) to (3) is no longer satisfied, then the supply rate of oxygen gas from the upper lance, the height of the upper lance, the height position of the movable bell, and the supply rate of the oxidizing gas or inert gas blown from the lower blow lance , return to the values before the adjustments, and after that the decarburization refining continues.

As a result of the implementation of the present invention, the emission frequency has been reduced to about 1/3 of the number of emissions prior to the implementation of the present invention.

EXAMPLE 2

Decarburization of molten pig iron 5 was carried out analogously to example 1, using the same converter equipment (top and bottom purge converter) as in example 1.

During the decarburization refining, the image of the throat combustion reaction flame 16 appearing in the gap between the converter mouth 2 and the movable bell was successively captured by the spectroscopic camera 6 with a predetermined measurement time interval Δt set to 1 second as in Example 1. The image analyzer 7 measured the spectrum emission (image data) from the acquired image, and performed analysis at selected emission wavelengths in the wavelength range of 580 to 620 nm in the acquired emission spectrum at each measurement point, and calculated an emission intensity index at each wavelength. The wavelength used here (ie the selected wavelength) was 610 nm. The analysis was performed by the method of linear analysis of image data on any scan line.

Calculations in accordance with the above formulas (1) - (3) were performed using the radiation intensity index obtained at a specific wavelength at each measurement moment. In this case, the values of the constants L_one, L₂, m₀, m_one and m₂ in formulas (1) - (3) are set the same as in example 1, but the threshold values for determining C₀, C_oneand C₂were defined as follows. The operating data obtained for 200 heats of the decarburization refining similar to that described in Example 1 was divided into four groups depending on the value of the average value of the oxygen gas supply rate during the oxygen purge, and threshold values were determined for each of the four groups. In other words, each threshold value in formulas (1) - (3) is set to four values depending on the value of the average value of the oxygen gas supply rate.

In real work, the average value of the flow rate of gaseous oxygen was calculated sequentially, and one of the four values of each threshold value was used, and it is determined depending on the average value of the rate of supply of gaseous oxygen. A prediction result indicating the occurrence of a metal outburst was determined when all formulas (1) to (3) were executed with the threshold values set as described above. When a prediction result indicating the occurrence of a metal outburst was determined, then, in accordance with the same standards as in Example 1, one, two or more of the following actions were performed: adjusting the supply rate of oxygen gas from the top blow lance, adjusting the height of the top blow lance , adjusting the position of the movable bell in height, adjusting the feed rate of the oxidizing gas or inert gas blown from the lower blow lance, and adding the soothing material to the converter. If any of the formulas (1) - (3) were no longer satisfied, then the rate of supply of gaseous oxygen from the upper blast lance, the height of the upper blast lance, the height position of the movable bell and the supply rate of the oxidizing gas or inert gas blown in were returned to the values before adjustments. from the lower blow lance, and thereafter the decarburization refining continued.

It was confirmed that due to the determination of the threshold values in the formulas (1) to (3) as described above, the occurrence of a metal burst during decarburization refining can be reduced to a level equal to or lower than that in Example 1, and decarburization refining can be stabilized. .

EXAMPLE 3

As described above, when determining the situation inside the converter based on the spectral analysis of the combustion reaction flame 16 in the neck, false detection may occur in some cases due to a change in the situation, for example, when the field of view is obstructed by the passing of a faucet or metal deposition on the neck. Therefore, the threshold values for determining C ₀ , C ₁ and C ₂ in formulas (1) - (3) it is desirable to change during the operation of the converter for each heat.

In view of the above, neural network-type machine learning was performed using, as training data, offline analysis data obtained from 2000 melts, each of which carried out a spectral analysis of the flame 16 of the combustion reaction in the neck. A total of 30 items, such as weight of molten pig iron, weight of scrap iron, temperature of molten pig iron before decarburization refining, amount of auxiliary raw materials added, oxygen supply rate (oxygen gas supply rate from the top blast lance) at the stage in the blowing process, gas flow rate injected from below, lance height, off-gas flow rate, off-gas composition, and movable bell height are taken as input, and the number of hidden layers is set to five.

The detection thresholds of C ₀ , C ₁ and C ₂ in formulas (1) to (3) were set by the machine learning threshold determination method as described above, and decarburization refining of the molten pig iron was performed in the same manner as in Example 1 , using the same converter equipment (converter with top and bottom blast) as in example 1. The values of the constants L ₁ , L ₂ , m ₀ , m ₁ and m ₂ in formulas (1) - (3) are set to the same same as in example 1.

During the whole time of the decarburization refining purge, the image of the combustion reaction flame 16 emerging from the mouth of the converter 2 was sequentially captured by the spectroscopic camera 6 with a predetermined measurement time interval Δt: 1 sec, as in Example 1. From the obtained image, the emission spectrum (image data) was measured and recorded by image analyzer 7.

A prediction result indicating the occurrence of a metal outburst was determined when all formulas (1) to (3) were executed. If a prediction result was determined indicating the occurrence of a metal outburst, then one, two or more of adjusting the flow rate of oxygen gas from the top blow lance, adjusting the height of the top blow lance, adjusting the height position of the movable bell, adjusting the flow rate of oxidizing gas or inert gas were performed. blown from the lower blow lance and adding a soothing material to the converter (as in Invention Example 3) according to the same standards as in Example 1.

To compare examples 1 to 3, the decarburization refining described in example 1, which was performed after pre-setting the detection thresholds C ₀ , C ₁ and C ₂ in formulas (1) - (3) to predetermined values (referred to as invention example 1) , and the decarburization refining described in Example 2, which was performed after setting the threshold values for determining C ₀ , C ₁ and C ₂ in formulas (1) to (3) depending on the value of the average value of the flow rate of oxygen gas during oxygen purge (called an example of the invention 2).

In each of the examples of the invention 1, 2 and 3 performed more than 100 melts. In each of the operations, when the prediction result indicating the occurrence of a metal outburst was determined, then, according to the same standards as in Example 1, one, two or more operations of adjusting the oxygen gas supply rate from the upper lance, adjusting the height upper lance, adjusting the height position of the movable bell, adjusting the flow rate of oxidizing gas or inert gas blown from the lower lance, and adding soothing material to the converter.

In addition, if any of formulas (1) to (3) is no longer satisfied, then the flow rate of oxygen gas from the upper lance, the height of the upper lance, the height position of the movable bell, and the flow rate of oxidizing gas or inert gas blown from the lower blow lance, returned to the values before the adjustments, and after that the decarburization refining continued.

Table 3 shows the results of operation in Invention Examples 1, 2, and 3. Table 3 further shows the result of prior art decarburization refining that was performed without predicting a metal release. The metal ejection frequency in Table 3 is given as a percentage representing the number of heats in which metal ejection occurred relative to the total number of heats (100 heats).

Table 3

Spectroscopic determination Plants based on the flow of gaseous oxygen Machine learning The appearance of a metal ejection
(%) Purge time (min) Amount of soothing material used (kg/t) Invention example 1

- - 16 19.3 1.7 Invention Example 2

- eleven 19.0 1.2 Invention Example 3

-

eight 18.9 0.7 State of the art - - - 39 19.6 2.3

As can be seen from Table 3, in Invention Example 3, the frequency of occurrence of the surge was low, the duration of the purge due to the occurrence of the surge was reduced, and the amount of soothing material used was also reduced.

List of symbols

1 converter hardware

1A converter equipment

2 converter

3 top blow lance

4 lower blow lance

5 liquid pig iron

6 spectroscopic camera

7 image analyzer

8 control computer

9 neck

10 movable bell

11 lance height adjuster

12 oxidizing gas flow regulator

13 gas flow regulator blown in from below

14 regulator for adding auxiliary raw materials

15 movable bell height adjuster

16 combustion reaction flame in the throat

17 oxidizing gas jet stream

18 pipe for supplying oxidizing gas to the upper blow lance

19 pipe for supplying cooling water to the upper blow tuyere

20 pipe for draining cooling water from the upper blow lance

21 machine learning computers

Claims (37)

1. A method for predicting the release of metal in a converter that occurs during decarburization refining in a converter in which molten pig iron is produced into liquid steel by blowing oxidizing gas into molten pig iron in the converter from an upper blast lance or blowing also with oxidizing gas or inert gas from a lower blast lance for performing decarburizing refining of molten pig iron, characterized in that measuring the emission spectrum of the flame of the combustion process in the neck coming out of the neck of the converter; calculate the emission intensity of the measured emission spectrum at a wavelength in the range of 580 - 620 nm and predict the occurrence of a metal release based on the change, in the time series, of the calculated radiation intensity, wherein said change, over time, in the irradiance intensity is determined according to a determination formula using a moving average, and as the said definition formula, the following formulas (1) to (3) are used:

where I(n, m ₀ ) denotes the moving average, in arbitrary units, of the radiation intensity index from the measurement point nm ₀ to the measurement point n; I(nL ₁ , m ₁ ) denotes the moving average, in arbitrary units, of the radiation intensity index from the measurement point nL ₁ -m ₁ to the measurement point nL ₁ ; I(n-2L ₁ , m ₁ ) denotes the moving average, in arbitrary units, of the radiation intensity index from the measurement point n-2L ₁ -m ₁ to the measurement point n-2L ₁ ; I(n, m ₂ ) denotes the moving average, in arbitrary units, of the radiation intensity index from the measurement point nm ₂ to the measurement point n; I(nL ₂ , m ₂ ) denotes the moving average, in arbitrary units, of the radiation intensity index from the measurement point nL ₂ -m ₂ to the measurement point nL ₂ ; C ₀ , C ₁ and C ₂ are detection thresholds and satisfy the relationships C ₀ >0, C ₂ >0 and C ₁ <C ₂ ; L ₁ and L ₂ are constants, each of which is an integer greater than or equal to 1; m ₀ , m ₁ and m ₂ are constants, each of which is an integer greater than or equal to 0, wherein when the conditions of all formulas (1) to (3) are met, it is determined that the prediction result indicates the occurrence of a metal outburst. 2. The method for predicting the release of metal in the converter according to claim 1, in which the threshold values for determining C ₀ , C ₁ and C ₂ in formulas (1) - (3) are determined using at least one of the following parameters: change in radiation intensity, off-gas flow rate, off-gas components, top blow lance oxygen feed rate, and top blow lance height during oxygen purge. 3. The converter metal release prediction method according to claim 1, wherein the detection threshold values of C ₀ , C ₁ and C ₂ in formulas (1) to (3) are determined by machine learning using at least one of the following parameters: change radiation intensity, off-gas flow rate, off-gas components, top lance oxygen feed rate, and top lance height during oxygen purge. 4. A method for the production of liquid steel from molten pig iron in a converter, characterized in that liquid steel is produced from molten pig iron, and using the method for predicting the release of metal in the converter according to any one of paragraphs. 1 to 3 determine the prediction result indicating the occurrence of a metal outburst, while after determining said prediction result indicating the occurrence of a metal outburst, at least one of the following adjustments is made: adjusting the flow rate of the oxidizing gas blown from the top blow lance, adjusting the height of the top blow lance, adjusting the position of the movable bell in height, adjusting the flow rate of the oxidizing gas or inert gas injected from the lower blow lance and the addition of a soothing material. 5. A system for predicting the release of metal in a converter occurring during decarburization refining in a converter in which liquid steel is produced from molten pig iron in the converter by blowing oxidizing gas into the molten pig iron in the converter from the upper blast lance or blowing also with oxidizing gas or inert gas from the lower blast lance to perform decarburization refining of molten pig iron, the system comprising: a spectroscopic camera located near the converter with the possibility of obtaining an image of the flame of the combustion process in the neck through the gap between the converter and the movable bell; and an image analyzer capable of recording image data, with the possibility of extracting them, sent from the spectroscopic camera, calculating from said image data the radiation intensity in the radiation spectrum at a wavelength in the range of 580 - 620 nm and predicting the metal emission based on the calculated change, in time series, radiation intensity, moreover, the image analyzer is configured to calculate the change, in a time series, of the intensity of the radiation in accordance with the determination formula, which uses a moving average, and the following formulas (1) - (3) are used as the indicated definition formula:

where I(n, m ₀ ) denotes the moving average, in arbitrary units, of the radiation intensity index from the measurement point nm ₀ to the measurement point n; I(nL ₁ , m ₁ ) denotes the moving average, in arbitrary units, of the radiation intensity index from the measurement point nL ₁ -m ₁ to the measurement point nL ₁ ; I(n-2L ₁ , m ₁ ) denotes the moving average, in arbitrary units, of the radiation intensity index from the measurement point n-2L ₁ -m ₁ to the measurement point n-2L ₁ ; I(n, m ₂ ) denotes the moving average, in arbitrary units, of the radiation intensity index from the measurement point nm ₂ to the measurement point n; I(nL ₂ , m ₂ ) denotes the moving average, in arbitrary units, of the radiation intensity index from the measurement point nL ₂ -m ₂ to the measurement point nL ₂ ; C ₀ , C ₁ and C ₂ are detection thresholds and satisfy the relationships C ₀ >0, C ₂ >0 and C ₁ <C ₂ ; L ₁ and L ₂ are constants, each of which is an integer greater than or equal to 1; m ₀ , m ₁ and m ₂ are constants, each of which is an integer greater than or equal to 0, wherein the image analyzer is configured to determine a prediction result indicating the occurrence of a metal ejection when the conditions of all of the above formulas (1) to (3) are met. 6. A converter metal ejection prediction system according to claim 5, wherein the image analyzer includes a machine learning model for determining said C ₀ , C ₁ and C ₂ threshold values in formulas (1) to (3) by machine learning with using at least one of the following parameters: change in radiation intensity, off-gas flow rate, off-gas components, oxygen gas supply rate from the top blow lance, and height of the top blow lance during oxygen purge. 7. The system for predicting the release of metal from the converter according to claim 5, also containing a machine learning computer that includes a machine learning model for determining the specified threshold values for determining C ₀ , C ₁ and C ₂ in formulas (1) - (3 ) through machine learning using at least one of the following parameters: change in emission rate, off-gas flow rate, off-gas components, oxygen gas supply rate from the top blow lance, and height of the top blow lance during oxygen purge.

Images