JP2016141828A - Blast furnace charging material detecting method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a blast furnace charging material detecting method capable of detecting the kind and moisture content of a charging material as a state.SOLUTION: A blast furnace charging material detecting method includes: in the case of using an arithmetic processing unit 7 and a database 8 and detecting a state of a blast furnace charging material, spectrally separating a reflected light in a near-infrared region from a charging material whose kind and moisture content are clear so as to acquire reflected spectrum information comprising spectrum for the number of M pieces of wavelength; and giving the kind and moisture content of the charging material as the property to the information and storing in the database 8 followed by repeating the process for a charging material having different kind and moisture content. Furthermore, the method includes: after newly acquiring reflection spectrum information from a charging material whose kind and moisture content are unclear, demanding a distance between the reflection spectrum information stored on an M-dimensional space, and newly acquired reflection spectrum information; selecting from the database 8, stored reflection spectrum information whose distance is small; and detecting the kind and moisture content of a charging material which newly acquires reflection spectrum from the property of the selected reflection spectrum information.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a blast furnace charge detection method for detecting the state of a charge charged in a blast furnace, and for example, how much coke, ore (iron ore), and sintered ore are charged in the blast furnace, respectively. In this state, the state in which the blast furnace is charged is detected.

  At present, it cannot be said that management of the amount and properties of the charge charged into the blast furnace from the top of the blast furnace, for example, the amount of moisture and the powder adhesion rate, is sufficient. For example, if the state of the charge charged into the blast furnace can be detected from the conveyor disposed above the top of the blast furnace, for example, the state of ventilation in the blast furnace can be managed, and the operation of the blast furnace is stabilized. It is also possible to do. As such a blast furnace charge detection apparatus, there exist some which are described in the following patent document 1 and patent document 2, for example. Among these, the blast furnace charge detection device described in Patent Document 1 arranges a coil sensor so as to surround the discharge port of the raw material storage layer, and the raw material is based on the output value of the coil sensor that changes as the raw material is discharged. The degree of mixing is detected. Moreover, the blast furnace charge detection apparatus described in the following Patent Document 2 performs image processing on an image captured by a camera and detects the particle size distribution of coke.

Japanese Patent No. 4802739 JP 2003-83868 A

However, the blast furnace charge detection device described in Patent Document 1 only detects the mixing degree of the charge, and the blast furnace charge detection device described in Patent Document 2 detects the coke particle size. Therefore, there is insufficient information on the charges to stabilize the blast furnace operation. Further, since the blast furnace charge detection device described in Patent Document 2 only detects the particle size of coke, not only coke but also ore and sintered ore are mixed and charged as in a blast furnace. In this case, since the image processing parameters for particle size detection are different, there is a possibility that the respective particle sizes cannot be detected properly.
The present invention has been made paying attention to the above problems, and an object of the present invention is to provide a blast furnace charge detection method capable of detecting at least the type of charge. It is.

  In order to solve the above problems, according to one aspect of the present invention, a database capable of storing a plurality of data together with attributes, an arithmetic processing function, and data stored in the database in contact with the database In the near-infrared region of the reflected light from the charge of a known type at least when detecting the state of the charge charged in the blast furnace using the processing function capable of reading And at least the type of the charge is assigned as an attribute to the reflection spectrum information and stored as data in the database, and this is repeated for at least different types of charges. A plurality of reflection spectrum information to which at least the kind of charge is given as an attribute is stored as data in a database, and at least Reflection spectrum information obtained by spectroscopically reflecting reflected light in the near-infrared region from a non-bright charge is newly acquired, and reflection spectrum information stored in the database and newly acquired reflection spectrum information data are set in advance. In the case of a spectrum corresponding to a number of wavelengths or frequencies, the distance between the stored reflection spectrum information in the M-dimensional space and the newly acquired reflection spectrum information is obtained, and the distance from the smallest or smallest is previously determined. Specified number of stored reflection spectrum information from the database is selected from the database, and blast furnace charging is performed to detect at least the type of charge from which the reflection spectrum information has been newly acquired from the selected reflection spectrum information attributes. An object detection method is provided.

  According to the present invention, reflected spectrum information is newly acquired by spectroscopically reflecting reflected light in the near-infrared region from the charge, and at least from the newly acquired reflection spectrum information and the reflected spectrum information stored in the database. The type of the charge can be detected as the state.

It is a schematic block diagram which shows one Embodiment of the blast furnace charge detection method of this invention. It is explanatory drawing of the reflection spectrum information from the charge acquired and normalized by the blast furnace charge detection method of FIG. 1 when a moisture content or a powder adhesion rate is constant. It is explanatory drawing of the reflection spectrum information from the ore actually acquired when the moisture content or the powder adhesion rate is constant. It is explanatory drawing of M-dimensional space which plotted the reflection spectrum information from the charge actually acquired when the moisture content or the powder adhesion rate was fixed. It is explanatory drawing of the reflectance spectrum information normalized from the ore when a moisture content changes. It is explanatory drawing of the normalized reflection spectrum information from coke when a moisture content changes. It is explanatory drawing of the normalized reflection spectrum information from a sintered ore when a moisture content changes. It is explanatory drawing which shows the correlation of the moisture content of a charging material, and a powder adhesion rate. It is explanatory drawing of the reflection spectrum information from the ore actually acquired when the moisture content changed. It is explanatory drawing of M-dimensional space which plotted the reflection spectrum information from the ore actually acquired when the moisture content changed. It is explanatory drawing of M-dimensional space which plotted the reflection spectrum information from the coke actually acquired when the moisture content changed. It is explanatory drawing of M-dimensional space which plotted the reflection spectrum information from the sintered ore actually acquired when the moisture content changed. It is explanatory drawing of the correct answer rate of the kind of charge detected with the blast furnace charge detection method of FIG. It is explanatory drawing which shows the correlation of the moisture content and the actual value of the charge detected with the blast furnace charge detection method of FIG.

  Next, an embodiment of the blast furnace charge detection method of the present invention will be described with reference to the drawings. FIG. 1 is a schematic configuration diagram of the blast furnace charge detection method of this embodiment. In the blast furnace 1, ore (iron ore), coke, and sintered ore are charged as charges from the top of the furnace, and hot air is blown from the tuyere 2 at the bottom of the furnace to the deposits in the furnace. The charge is reduced, i.e., ironmaking. The hot metal that has flowed down to the hearth part is discharged together with the hot metal from a tap outlet formed on the outer periphery of the bottom of the blast furnace 2.

  In the present embodiment, the conveyor 3 is disposed above the top of the blast furnace 1 in order to charge charges such as ore, coke, and sintered ore from the top of the blast furnace 1. The charge conveyed by the conveyor 3 is once charged into a hopper (also referred to as a bunker) 4, and is then uniformly charged into the furnace through the rotary distributor 5. In the present embodiment, a camera 6 is disposed above the conveyor 3, and the type of the charge and the amount of moisture are detected from the reflected light of the charge on the conveyor 3 captured by the camera 6. In this embodiment, illumination is necessary to obtain a stable light amount, but illumination in a preset wavelength region may be used as necessary.

  The camera 6 employed in this embodiment is not a general camera that captures an imaging region that expands two-dimensionally, but is a so-called line sensor type spectroscopic camera. This line sensor is configured by arranging image sensors composed of CCD, CMOS, etc. one pixel at a time in a one-dimensional direction, and each image sensor detects the color and brightness of the reflected light from the charge. The camera 6 constituted by this line sensor, for example, splits the reflected light in the near infrared region for each pixel out of the charged material reflected light one-dimensionally in the direction orthogonal to the conveying direction of the conveyor 3. Thus, the reflection spectrum (reflection intensity) for each wavelength (may be for each frequency) can be acquired as reflection spectrum information.

  The reflection spectrum of the charge obtained by spectrally dividing the reflected light in the near-infrared region in this way is taken into an arithmetic processing device 7 having an advanced arithmetic processing function such as a computer, for example. Detection of the type of the input, the moisture content of the input from which the type has been detected, or the powder adhesion rate of the input from which the type has been detected is detected. The powder adhesion rate is also referred to as the powder rate. This arithmetic processing unit 7 is connected to a database 8. The database 8 is a storage area of a large-capacity storage device, and in this embodiment, the arithmetic processing device 7 can store the reflection spectrum information captured by the arithmetic processing device 7 with an attribute. Further, the arithmetic processing unit 7 can contact the database 8 and read the reflection spectrum information stored in the database 8 together with the attribute.

  The arithmetic processing unit 7 includes an input / output unit, an arithmetic processing unit, a storage unit, and the like, like an existing computer. The blast furnace charge detection method of this embodiment is constructed by software that is stored and executed in the arithmetic processing unit 7, and an arithmetic processing unit 7 and a database 8 that function with the software. FIG. 2 shows a wavelength reflection spectrum of the charge reflected light for each type. In infrared spectroscopy, when the surface state of a detected object is complicated like a blast furnace charge, the resulting reflection spectrum varies. Thus, for example, FIG. 2 shows a sample in which a large number of reflection spectra are sampled and stored in the database 8 for each type of charge, and the dispersion is normalized. In the drawing, the moisture content (or powder adhesion rate) of each charge is constant and does not change, for example, the dry condition where the moisture content is less than 1 wt%. For example, the ore looks reddish brown, and the appearance of coke and sintered ore is gray. Therefore, the pattern of the ore reflection spectrum is characteristic to the pattern of the reflection spectrum of coke and sintered ore. On the other hand, the pattern of the reflection spectrum of coke is similar to the pattern of the reflection spectrum of sintered ore. However, there is a clear pattern difference between them, particularly in the wavelength region of 1000 to 1200 nm. By comparing the reflection spectra for each wavelength, it is possible to detect, for example, the type of the newly obtained charge as the state of the charge.

  As described above, the reflection spectrum of the charge actually obtained varies as shown in FIG. 3 even if the moisture content or the powder adhesion rate is constant. In the figure, a plurality of reflection spectra of ores under the above-mentioned dry conditions are acquired and expressed in terms of wavelength. Since ore, coke and sintered ore are natural products or those obtained by primary processing of natural products, they have surface irregularities and the surface state is not uniform. Therefore, the reflection spectrum also varies.

  Therefore, in this embodiment, a plurality of wavelengths that well represent the characteristics of the reflection spectrum of each charge are set in advance, and a large number of reflection spectra of those wavelengths are acquired and expressed in spatial coordinates as reflection spectrum information. For example, when the number of wavelengths set in advance with respect to the reflection spectrum is M, an M-dimensional space is configured by representing each of these wavelengths as an axis, and a number of reflections acquired in the M-dimensional space. Represents spectral information. Each piece of acquired reflection spectrum information appears as a point in the M-dimensional space. FIG. 4 schematically shows an M-dimensional space in which the reflection spectrum information from the charge actually obtained is plotted when the moisture content (or the powder adhesion rate) is dry and constant. As is apparent from the figure, the reflection spectrum information is gathered even though there are variations for each type of charge, that is, ore (◯), coke (Δ), and sintered ore (□). Therefore, for example, when the reflection spectrum information of the charge as indicated by a star (☆) in FIG. 3 is newly acquired, the reflection spectrum information group closest to the reflection spectrum information is newly added to the reflection spectrum information. It turns out that it seems to be the kind of the charge which was acquired. In addition to the wavelength that characterizes the type of charge, the wavelength that characterizes the moisture content (or powder adhesion rate) of the charge may be selected for the M wavelengths set in advance, as described later. .

In the reflection spectrum information of the charge obtained by sampling, for example, the type of the charge is given as an attribute and stored in the database 8. When the arithmetic processing unit 7 newly acquires the reflection spectrum information from the camera 6, the M dimension of the newly acquired reflection spectrum information and the stored reflection spectrum information is obtained for all the reflection spectrum information stored in the database 8. The distance d in space is obtained. The stored reflection spectrum information having the smallest distance d is selected from the database 8, and the type of the charge from which the reflection spectrum information is newly acquired can be detected as the state from the attribute of the selected reflection spectrum information. The distance d between the newly acquired reflection spectrum information and the stored reflection spectrum information in the M-dimensional space is the reflection spectrum of each wavelength of the newly acquired reflection spectrum information as λ _obji (i = 1, 2,... , M), where the reflection spectrum of each wavelength of the stored reflection spectrum information is λ _dbi (i = 1, 2,..., M), the following equation 1 is obtained.
d = √ ((λ _obj1 −λ _db1 ) ² + (λ _obj2 −λ _db2 ) ² +… + (λ _objM −λ _dbM ) ² ) (1)

  Alternatively, for newly acquired reflection spectrum information, instead of the stored reflection spectrum information having the smallest distance d, the number of reflections stored in advance from the smallest distance d, for example, K stored reflections. Spectral information may be selected from the database 8, and the type of the charge from which the reflection spectrum information is newly acquired may be detected as a state from the attribute of the reflection spectrum information. For example, even if the attribute of the reflection spectrum information with the smallest distance d is “ore”, if the attribute of the remaining “K−1” pieces of reflection spectrum information is “sintered ore”, the reflection is newly performed. The charge from which the spectral information has been acquired may be “sintered ore”. Further, when the state of the charge is a continuous value such as a moisture amount and a powder adhesion rate described later, for example, the amount of moisture and the powder adhesion rate as attribute information of the selected K reflection spectrum information. It is also possible to detect the moisture content and the powder adhesion rate of the charge from which the reflection spectrum information is newly acquired by using a weighted average value in which the weight is increased as the average value and the distance d are smaller.

  As described above, the amount of moisture can be given as an attribute of the reflection spectrum information of the charge. FIG. 5 shows the reflectance spectrum (reflectance in the figure) of the ore, FIG. 6 of the coke, and FIG. 7 of the sintered ore, with the variation normalized when the water content is changed. In either case, a change in the reflection spectrum is observed with a change in the amount of water in the water absorption wavelength band near 1450 nm and 1950 nm. Therefore, the wavelengths including these wavelength bands are included in the M wavelengths described above, a large number of reflection spectra with different moisture amounts are sampled for each type of charge, and the type and amount of moisture are used as attributes. The assigned reflection spectrum information is stored as data in the database 8, and the newly obtained reflection spectrum information of the charge is compared with the reflection spectrum information for each moisture amount, and the moisture content of the charge is detected as a state. can do. Incidentally, the moisture content of each charge and the powder adhesion rate of each charge are in a linear relationship with each other as shown in FIG. That is, as the amount of water in the charge increases, the amount of powder adhering to the charge increases accordingly, and the powder adhesion rate of the charge also increases. Therefore, the water content obtained from the reflection spectrum information may be considered as the powder adhesion rate.

  Even when the amount of moisture (or powder adhesion rate) changes in this way, there is variation in the reflection spectrum that is actually obtained from the charge with the same amount of moisture. FIG. 9 shows the reflection spectrum actually obtained when the moisture content of the ore changes. However, the reflection spectrum actually obtained when the moisture content also changes for coke and sintered ore. Exists. The reflection spectrum information actually obtained when the water content of the ore represented as shown in FIG. 5 is changed by normalization is schematically shown in the M-dimensional space of FIG. Similarly, the reflection spectrum information actually obtained when the moisture content of the coke expressed as shown in FIG. 6 is changed by normalization is schematically shown in the M-dimensional space of FIG. Moreover, the reflection spectrum information actually acquired when the moisture content of the sintered ore expressed as shown in FIG. 7 is changed by normalization is schematically shown in the M-dimensional space of FIG. As can be seen from the fact that the space is M-dimensional, these are data of reflection spectra of M wavelengths including wavelengths in a wavelength region in which a change in moisture content (or powder adhesion rate) is noticeable.

  The distance d between the reflection spectrum information actually acquired and stored in the database 8 when the water content changes in this way and the newly acquired reflection spectrum information of the charge are calculated in the same manner as described above, and the distance d Of the stored reflection spectrum information of the smallest or the smallest number of the distance d, for example, K pieces of stored reflection spectrum information are selected from the database 8, and the selected reflection spectrum information of the selected reflection spectrum information is selected. The type and moisture content (or powder adhesion rate) of the newly acquired reflection spectrum information from the attribute can be detected as the state. Since the moisture content (or powder adhesion rate) is a continuous value, the weight value increases as the average value of the moisture content (or powder adhesion rate) or the distance d becomes smaller as the attribute information of the K reflection spectrum information selected. If the weighted average value or the like is used, the moisture content (or powder adhesion rate) of the charge can be detected more accurately.

  As is clear from FIGS. 10 to 12, the direction of the change in the reflectance spectrum information when the water content (or the powder adhesion rate) changes in each of the ore, coke, and sintered ore is roughly determined. Yes. Therefore, by performing principal component analysis of reflection spectrum information when the amount of water changes for each charge, and performing dimensional compression of the data, it is possible to detect a state more resistant to errors. As shown in FIGS. 10 to 12, dimensional compression by principal component analysis is, for example, a projection onto an axis (= principal component) that maximizes dispersion accompanying changes in the amount of moisture (or powder adhesion rate). The reflection spectrum information, which varies in the vicinity of, can be projected onto the main component to perform dimensional compression. Since the dimensionally compressed reflection spectrum information reflects only the change in moisture content (or powder adhesion rate), this dimensionally compressed reflection spectrum information is compared with the newly acquired reflection spectrum information. By doing so, it becomes possible to detect the amount of moisture (or powder adhesion rate) more accurately.

  In the blast furnace, the moisture content and the powder adhesion rate of the charge affect the air permeability of the blast furnace. Specifically, the greater the amount of moisture and the greater the powder adhesion rate, the lower the furnace air permeability. A decrease in furnace air permeability leads to destabilization of blast furnace operation, so grasping the state of these charges enables stabilization of blast furnace operation. In the blast furnace charge detection method of this embodiment, not only the type of blast furnace charge, but also the water content and powder adhesion rate of those charges can be detected, so that the blast furnace operation is stable. Can be achieved. In FIG. 13, the correct answer rate of the kind of charging detected with the blast furnace charging detection method of this embodiment is shown. As described above, since the pattern of the reflection spectrum of the ore has distinct characteristics, the accuracy rate of the discrimination was 100%. On the other hand, although the reflection spectrum pattern is similar between coke and sintered ore, since there is a difference in the change rate with respect to the wavelength in the wavelength range of 1000 to 1200 nm as described above, 90% although there is a slight error. The accuracy rate exceeded. FIG. 14 shows the correlation between the water content of the charge detected by the blast furnace charge detection method of this embodiment and the actual value. As is clear from the figure, the moisture content can be detected although there is variation.

  In the blast furnace charge detection method of this embodiment, the type of the charge, the amount of water, and the powder adhesion rate are detected as the state from the reflection spectrum information, but the state of the charge is determined as the state of the charge. Ingredient amounts can also be added. The amount of component of the charge is reflected in the reflection spectrum information of the charge. Therefore, for example, if the reflection spectrum information is added to the reflection spectrum information as an attribute and the component amount of the charge affecting the operation of the pig iron and the operation of the blast furnace, and it is stored as data in the database, the newly acquired reflection spectrum information It is possible to detect, as a state, the component amount of the charge from which the reflection spectrum information is newly obtained from the attribute of the data of the stored reflection spectrum information similar to.

  As described above, in the blast furnace charge detection method of the present embodiment, the database 8 capable of storing a plurality of data together with the attributes, the arithmetic processing function, and the data stored in the database in contact with the database are stored. Using the arithmetic processing function 7 that can be read together with the attribute, the state of the charge charged in the blast furnace is detected. At that time, at least a reflection spectrum information obtained by dispersing the reflected light in the near-infrared region from a charge of which the type is clear is obtained, and at least the type of the charge is assigned as an attribute to the reflection spectrum information to create a database. 8 is stored as data, and this is repeated for at least different types of charges, and a plurality of reflection spectrum information to which at least the types of charges are given as attributes are stored in the database 8 as data. In that state, a reflection spectrum obtained by spectrally dividing the reflected light in the near infrared region from at least a kind of unclear charge is newly acquired, and the reflection spectrum information stored in the database 8 and the newly acquired reflection spectrum information are obtained. When the data is a spectrum of M wavelengths or frequencies set in advance, the distance between the stored reflection spectrum information in the M-dimensional space and the newly acquired reflection spectrum information is obtained. Then, stored reflection spectrum information corresponding to a preset number from the smallest or smallest distance is selected from the database 8, and a reflection spectrum is newly acquired from the attribute of the selected reflection spectrum information. At least the type of the entry is detected as a state. As a result, at least the type of the charge can be detected with high accuracy.

In addition to the type of charge, at least one of the moisture content of the charge and the powder adhesion rate of the charge is given as an attribute of the reflection spectrum information, and a new reflection is made from the selected reflection spectrum attribute. In addition to the type of charge from which the spectrum was acquired, at least one of the moisture content and the powder adhesion rate is detected. Therefore, the type of charge can be specified and the moisture content and powder adhesion rate of that type of charge can be detected. Based on these, for example, the blast furnace operation can be controlled by managing the blast furnace air permeability. It becomes possible to stabilize.
Further, by applying dimension compression by applying principal component analysis to at least one of the moisture content and the powder adhesion rate of the charge, it is possible to detect the state of the moisture content and the powder adhesion rate that are resistant to errors.

1 Blast Furnace 2 Tuyere 3 Conveyor 4 Hopper 5 Distributor 6 Camera 7 Arithmetic Processor 8 Database

Claims (3)

A database capable of storing multiple data with attributes;
With an arithmetic processing function, using an arithmetic processing function capable of reading the data stored in the database together with the attribute in contact with the database,
A blast furnace charge detection method for detecting a state of a charge charged in a blast furnace,
Reflection spectrum information obtained by spectroscopically reflecting reflected light in the near-infrared region from a charge of which the type is clear is obtained, and at least the type of the charge is assigned as an attribute to the reflection spectrum information and data is stored in the database. And repeatedly performing this for at least different types of charges, and storing at least a plurality of reflection spectrum information assigned with the types of charges as attributes in the database as data,
Newly obtained reflection spectrum information obtained by spectroscopically reflecting the reflected light in the near infrared region from at least a kind of non-lightening charge,
When the reflection spectrum information stored in the database and the data of the newly acquired reflection spectrum information are spectra for M wavelengths or frequencies set in advance, the reflection spectrum stored in the M-dimensional space is stored. The distance between the information and the newly acquired reflection spectrum information is obtained, and the stored reflection spectrum information corresponding to a preset number is selected from the database from the smallest or smallest distance, and the selected reflection is obtained. A method for detecting a blast furnace charge, comprising detecting at least a type of a charge for which reflection spectrum information is newly acquired from an attribute of spectrum information as a state.   As the attribute of the reflection spectrum information, in addition to the type of the charge, at least one of the moisture content of the charge and the powder adhesion rate of the charge is given, and from the attribute of the selected reflection spectrum information 2. The blast furnace charge detection method according to claim 1, wherein at least one of a moisture content and a powder adhesion rate is detected in addition to the kind of charge from which reflection spectrum information is newly acquired. 3. The blast furnace charge detection method according to claim 2, wherein dimension compression is performed by applying principal component analysis to at least one of the moisture content and the powder adhesion rate of the charge.

Images