Composition identification device and control system

Abstract

To provide a burning determination apparatus capable of making burning determination of an object in smelting, and a control system.SOLUTION: A burning determination apparatus includes an acquisition part, an extraction part, a calculation part, and a classification part. The acquisition part acquires, as one pair of object image data, each of a plurality of images obtained by imaging a process in which a mineral concentrate including recovered metals burns by being heated, and time information about the plurality of images. The extraction part extracts a pixel value of an image included in the object image data. The calculation part calculates a probability of similarity to a burning pattern indicating a burning state of the mineral concentrate from the extracted pixel value and the time information included in the object image data. The classification part classifies the burning pattern on the basis of the probability.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a combustion determining device, a composition specifying device, and a control system. [Background technology]

[0002] Conventionally, in order to reduce the loss of recovered metals in slag in metal smelting, various factors have been investigated, such as physical factors such as viscosity related to phase separation between slag and matte, the influence of magnetite in slag, and the origin of the supplied concentrate.In particular, the supplied concentrate has a variety of compositions depending on the origin, making it difficult to control the concentrate treatment process while taking into account the origin of the supplied concentrate.

[0003] Therefore, for example, Patent Document 1 discloses a combustion assessment device that reduces the time delay to a level that does not pose a practical problem. The brightness and color of the flame change depending on the combustion state, and these changes correlate with the combustion state. The combustion assessment device in Patent Document 1 determines the degree of conformance to a typical combustion state using characteristic parameters calculated from the chromaticity coordinates of a specific color system that has a correlation. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 08-049845 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the technology of Patent Document 1 is not suitable for short-term combustion determination and composition identification of an object, such as the reaction of an object such as a concentrate that occurs in a smelting apparatus (flash furnace) used for metal smelting, etc.

[0006] An object of one aspect of the present invention is to provide a combustion determination device that can determine combustion of an object in smelting. Another aspect of the present invention aims to provide a composition identification device that can obtain the composition ratio of an object based on the combustion state from an image. Another aspect of the present invention is to provide a control system capable of controlling a smelting plant by combustion determination or composition identification. [Means for solving the problem]

[0007] The combustion determination device according to the embodiment includes an acquisition unit, an extraction unit, a calculation unit, and a classification unit. The acquisition unit acquires a set of target image data, each of which includes multiple images of a process in which a concentrate containing recovered metals is heated and combusted, and time information for each of the multiple images. The extraction unit extracts pixel values ​​of the images included in the target image data. The calculation unit calculates a probability representing the similarity between the extracted pixel values ​​and a combustion pattern representing the combustion state of the concentrate, based on the extracted pixel values ​​and the time information included in the target image data. The classification unit classifies the combustion patterns based on the probability.

[0008] The composition identification device according to the embodiment includes an acquisition unit, an extraction unit, a calculation unit, and an identification unit. The acquisition unit acquires, as a set of target image data, each of a plurality of images captured during the heating and combustion of a concentrate containing recovered metals and time information for each of the plurality of images. The extraction unit extracts pixel values ​​of the images included in the target image data. The calculation unit calculates a probability representing the similarity between the extracted pixel values ​​and the time information included in the target image data and the composition ratio of the concentrate. The identification unit identifies the composition ratio of the concentrate based on the probability. [Effects of the Invention]

[0009] One aspect of the embodiment can provide a combustion determination device that can determine combustion of an object in smelting. Another aspect of the embodiment can provide a composition identification device that can obtain a composition ratio of an object based on a combustion state from an image. Another aspect of the embodiment may provide a control system that can control a smelter by combustion determination or composition identification. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a schematic diagram of a flash melting furnace. [Figure 2] 1 is a diagram showing the configuration of a control system including a combustion determination device according to a first embodiment, and the functional configuration of the combustion determination device. [Figure 3] FIG. 10 is a diagram showing an example of temperature change with respect to combustion time for concentrates with different compositions. [Figure 4] FIG. 10 is a diagram showing an example of an image of concentrates with different compositions being combusted. [Figure 5] 1 is a flowchart showing a method for determining combustion of an object in smelting. [Figure 6] FIG. 10 is a diagram showing the configuration of a control system including a composition specifying device according to a second embodiment, and the functional configuration of the composition specifying device. [Figure 7] FIG. 2 is a block diagram illustrating an example of a hardware configuration of an information processing device. DETAILED DESCRIPTION OF THE INVENTION

[0011] For example, in flash smelting furnaces used in copper smelting, the separation of matte particles in the slag due to settling is promoted by the growth of the matte particle size. The matte particles of the supplied fine concentrate grow by colliding with each other, but fine matte particles of several tens of micrometers melt in a short time, so the molten matte particles grow by colliding with each other. Therefore, it is very important to understand the combustion state of the concentrate.

[0012] In addition, ores containing the target metals are processed into concentrates for smelting, but the content of substances other than the target metals varies depending on the place of origin and the mineral composition, which results in different combustion conditions. For this reason, it is also important to understand the changes in the composition ratios in the concentrate.

[0013] Hereinafter, a combustion determination device that enables combustion determination of an object in smelting, a composition identification device that enables identification of the composition ratio of an object, and a control system that enables control of a smelting apparatus according to embodiments of the present invention will be described. Note that the configurations shown in the following embodiments are merely examples, and the present embodiments are not limited to the illustrated configurations. Furthermore, to facilitate understanding of the description, the same components are assigned the same reference numerals in each drawing, and duplicated descriptions will be omitted.

[0014] In this embodiment, the target material is a raw material that is reacted by combustion or the like in a smelting apparatus (flash furnace), and examples thereof include concentrates.

[0015] <Smelting equipment (self-smelting furnace)> First, a smelting apparatus (flash smelting furnace) to which a control system including a combustion determination device or a composition identification device according to this embodiment is applied will be described. FIG. 1 is a schematic diagram of a flash smelting furnace. As shown in FIG. 1, the flash smelting furnace 1 includes a reaction shaft 2 and a concentrate burner 3. The reaction shaft 2 is a hollow structure. The concentrate burner 3 is provided at the top of the reaction shaft 2. In the flash smelting furnace 1, sulfur content in concentrate particles P supplied together with gas from the concentrate burner 3 burns below the concentrate burner 3, i.e., in the space within the reaction shaft 2, and the heat melts the concentrate particles P to form matte and slag.

[0016] First Embodiment Fig. 2 is a diagram showing the configuration of a control system including the combustion determination device according to the first embodiment, and the functional configuration of the combustion determination device. As shown in Fig. 2, the control system 100A includes a supply device 10, a camera 20 which is an imaging device, a combustion device 30, and a combustion determination device 40, which are connected to each other so as to be able to communicate with each other in any manner.

[0017] The supply device 10 supplies the concentrate after sorting and crushing the ore in advance in the ore dressing process. The concentrate to be supplied is stored in a storage device such as a hopper and is continuously transported, for example, in fixed amounts at a time. The transport device can be, for example, a powder transport device such as a screw conveyor, belt conveyor, or chain conveyor.

[0018] The camera 20 includes any sensor capable of photographing and capturing images. The camera 20 may be, for example, a web camera connectable to a USB (Universal Serial Bus) or an image sensor used in a digital camera. The camera 20 may also use, for example, a CCD (Charge Coupled Device), or a specialized sensor such as an infrared sensor or laser sensor, or may have the ability to visualize sensed data. Furthermore, the camera 20 may use grayscale information if the image is monochrome, or may acquire, for example, RGB information from a color image. Alternatively, one or more spectra (wavelength components) separated using a monochromator may be used.

[0019] The camera 20 is an imaging device that acquires image data of the combustion process. The camera 20 may acquire frames (still images) at predetermined intervals and send a predetermined number of frames to the combustion determination device 40. The image data acquired by the camera 20 may be stored in an image storage unit 47 within the combustion determination device 40. The image data includes at least one of still image data and moving image data.

[0020] The camera 20 may capture images through an observation window that can capture images of the inside of the flash smelting furnace 1 shown in Fig. 1. For example, a method may be used in which an observation window for monitoring is provided in a tuyere or the like that blows air into the flash smelting furnace 1, and the camera 20 is additionally installed in the observation window.

[0021] The combustion device 30 may be, for example, a burner connected to the reaction shaft 2 shown in FIG. 1. An auxiliary fuel burner for solid fuel may also be used as the burner. For example, when the combustion device 30 is an auxiliary fuel burner, raw materials containing solid sulfide are supplied from the outer periphery of the auxiliary fuel burner into the flash smelting furnace 1, and a reactive gas is also injected into the flash smelting furnace 1. The mixed gas containing the raw materials and the reactive gas comes into contact with the reactive gas flow adjusted to a predetermined speed, and the temperature inside the furnace rises due to the heat of the fuel, the sensible heat of the reactive gas, or radiant heat from the furnace inner wall, thereby causing the reaction of the concentrate to proceed.

[0022] The combustion determination device 40 determines the combustion state of the concentrate using a concentrate combustion determination classification model (hereinafter simply referred to as the model) that is learned using image data obtained from the camera 20. As shown in FIG. 2, the combustion determination device 40 includes an acquisition unit 41, an extraction unit 42, a learning unit 43A, a calculation unit 44A, a classification unit 45, a control unit 46, and an image storage unit 47. Each processing unit of the combustion determination device 40 will be described below. Note that each processing unit is realized by a processor 210 (see FIG. 7) of an information processing device 200 (see FIG. 7), which will be described later, reading and executing a program stored in a memory or the like.

[0023] As shown in Figure 2, the acquisition unit 41 acquires, as a set of target image data, each of a plurality of images captured during the process of a concentrate containing recovered metals being heated and burned, and time information for each of the plurality of images.

[0024] The acquisition unit 41 acquires image data of changes in the combustion of the concentrate captured for a predetermined period of time from the camera 20. If the camera 20 is an imaging device that captures still images at predetermined intervals, the acquired images are treated as a set of images along with time information in a chronological order.

[0025] The image may be data with a number of pixels and frame intervals appropriately set by the user depending on, for example, the type of target metal, the type of auxiliary fuel and reactive gas used in combustion, the combustion speed, etc.

[0026] The extraction unit 42 extracts pixel values ​​of the image included in the set of target image data. The extraction unit 42 extracts pixel values ​​of the image included in the set of target image data obtained from the acquisition unit 41, and sets the extracted pixel values ​​as feature quantities that indicate the features of the image to be processed.

[0027] The feature may be, for example, pixel values, which are color information, such as RGB value (256 levels) data, or an average value of the RGB values ​​of each pixel, or lightness, luminance, etc. For example, when specifying a composition value, it is preferable to use RGB value data, since this allows for detailed correspondence between color changes due to composition. Furthermore, instead of RGB value data, color system data such as CMY or CMYK may be used.

[0028] The feature amount may also be an image extracted by cutting out a portion from the image to be processed. For example, if the combustion position of the concentrate to be processed is roughly determined, the target area may be determined by specifying a predetermined area using a rectangle or the like. In this embodiment, since the target is a plurality of images arranged in chronological order, limiting the target area is preferable because it shortens the processing time. Note that the size of the image used for learning the predetermined area and the size used for processing by the combustion determination device 40 do not necessarily have to be the same.

[0029] The learning unit 43A may learn a combustion pattern in advance using a set of target image data acquired by the acquisition unit 41. The learning unit 43A calculates the occurrence probability of a combustion pattern representing the combustion state of the concentrate from pixel values ​​and time information (time series data) of each target image containing the pixel values. The learning unit 43A may calculate the occurrence probability of a combustion pattern representing the combustion state using a model previously learned by the learning unit 43A.

[0030] For example, if the concentrate is copper concentrate, when the pulverized concentrate is supplied from the supply device 10 and heated by the combustion device 30, the copper sulfides oxidize and generate heat. Figure 3 shows an example of temperature change over combustion time for copper concentrates of different compositions. Figure 3(a) shows a case where the first and second combustions of the copper concentrate are good, while Figure 3(b) shows a case where the first and second combustions of the copper concentrate are insufficient. As shown in Figure 3(a), when the first and second combustions of the copper concentrate are good, the temperature change has a steep peak. On the other hand, when the first and second combustions of the copper concentrate are insufficient, the combustion is slow, and the temperature change has a gentle peak.

[0031] Figure 4 shows examples of images of copper concentrates with different compositions being burned. Figure 4(a) corresponds to Figure 3(a) and shows a case where the first and second combustions of the copper concentrate were successful. Figure 4(b) corresponds to Figure 3(b) and shows a case where the first and second combustions of the copper concentrate were insufficient. As shown in Figure 4(a), if the first and second combustions of the copper concentrate were successful, the copper concentrate appears bright. On the other hand, as shown in Figure 4(b), if the first and second combustions of the copper concentrate were insufficient, the copper concentrate appears dark.

[0032] In learning the combustion pattern, the learning unit 43A may assign a teacher label to classify the temperature change shown in FIG. 3(a) as "good" and the temperature change shown in FIG. 3(b) as "insufficient," and then perform learning. These teacher labels can be changed as appropriate by the user. For example, the learning unit 43A may further assign a classification such as "other" to the teacher labels for learning.

[0033] Here, a learning model obtained by learning in the learning unit 43A will be described. The learning model for determining combustion patterns may be configured, for example, as a convolutional neural network (CNN). CNN is a type of machine learning algorithm that can be trained by supervised learning. Supervised learning is a type of machine learning in which a set of training data is provided to infer a learning model.

[0034] Each individual sample of training data is a combination of a data set (e.g., multiple images) and a desired output value or data set. A supervised learning algorithm analyzes the training data to generate a prediction function. Once derived through training, the prediction function can reasonably predict or estimate the correct output value or data set for a valid input. The prediction function can be formulated based on various machine learning models, algorithms, processes, etc.

[0035] The architecture of a CNN model includes multiple layers that convert inputs into outputs. Examples of layers include convolution layers, nonlinear operator layers that use activation functions, pooling layers, and affine (connection) layers. Each layer connects the previous and subsequent layers. The layer into which a dataset is input is sometimes called the input layer, and the layer that finally outputs the data is sometimes called the output layer. For example, the ReLu function, sigmoid function, and hyperbolic tangent function (tanh function) can be used as activation functions.

[0036] Furthermore, the learning unit 43A may learn a model using back propagation, stochastic gradient descent, or the like, based on a loss function such as squared error or cross entropy error.

[0037] The learning unit 43A may have an increased number of different layers. Examples of deep CNN models include AlexNet, VGGNet, GoogLeNet, and ResNet. The learning unit 43A is not limited to CNN, and may be any model that can use pixel values ​​and time information, such as a recurrent neural network (RNN) or a long short-term memory (LSTM).

[0038] The calculation unit 44A calculates the probability representing the similarity between the extracted pixel value and the combustion pattern representing the combustion state of the concentrate from the extracted pixel value and the time information included in the target image data. That is, the calculation unit 44A calculates the probability representing the similarity between a predetermined combustion pattern, input data (input data), and each pattern.

[0039] In this embodiment, the calculation unit 44A uses the learning model learned by the learning unit 43A to calculate the probability of similarity between the extracted pixel value and the combustion pattern representing the combustion state of the concentrate based on the time information contained in the target image data.

[0040] The classification unit 45 classifies the input data into combustion patterns based on the probabilities calculated by the calculation unit 44 A. That is, the classification unit 45 identifies which of the predetermined combustion patterns the input data corresponds to based on the probabilities calculated by the calculation unit 44 A.

[0041] The classification unit 45 sends the identified classification results to the control unit 46. The classification results regarding the combustion pattern obtained by the classification unit 45 are used via the control unit 46 to control the amount of concentrate supplied to the supply device 10 and the reaction gas and burner in the combustion device 30.

[0042] It is preferable to learn the model in advance, and therefore the model generated in advance by the learning unit 43A may be used in the combustion determining device 40.

[0043] The control unit 46 may change the settings of each processing unit, such as the amount of concentrate supplied by the supply device 10, control of the camera 20, and control of the reaction gas and burner in the combustion device 30. Control of the camera 20 includes, for example, the timing of image capture by the camera 20.

[0044] The image storage unit 47 may store original images acquired by the camera 20, or may store images after identification processing by the combustion determination device 40. For example, original images acquired via communication from an external terminal connected to a network may be stored in the image storage unit 47. Alternatively, an external storage device such as a USB memory or a recording medium such as a DVD (Digital Versatile Disk) storing original images may be connected to the control system 100A, and the original images may be read from the external storage device or recording medium and stored in the image storage unit 47.

[0045] The combustion determination device 40 may include a communication unit (not shown). This allows the control unit 46 to exchange various information, including images, with external devices connected to a network. Here, the network refers to communication resources for transmitting various information and includes various transmission paths connected by wire or wireless. The network includes, for example, communication paths such as the widely used Internet network, communication paths for mobile devices such as PHS (Personal Handyphone System), 4G (4th Generation), 5G (5th Generation), and LTE (Long Term Evolution), terrestrial broadcasting networks, satellite broadcasting networks, cable transmission networks, radio communication, millimeter wave communication, and radar communication. The control unit 46 can receive control commands from external devices connected to the network and change the settings of each processing unit in response to the control commands.

[0046] Fig. 5 is a flowchart showing a method for determining combustion of an object during smelting. As shown in Fig. 5, the combustion determination device 40 acquires, by the acquisition unit 41, a set of object image data, each of a plurality of images taken during the process of heating and burning a concentrate containing recovered metals, and time information for each of the plurality of images (acquisition step: step S11).

[0047] Next, the combustion determining device 40 extracts pixel values ​​of the image included in the set of target image data by the extracting unit 42 (extracting step: step S12).

[0048] Next, the combustion determining device 40 causes the learning unit 43A to learn a combustion pattern in advance using the set of target image data acquired by the acquisition unit 41 (learning step: step S13).

[0049] Next, the combustion determination device 40 calculates, by the calculation unit 44A, a probability representing the similarity with the combustion pattern representing the combustion state of the concentrate from the extracted pixel value and the time information included in the target image data (calculation step: step S14).

[0050] Next, the classification unit 45 classifies the combustion patterns based on the probability (classification step: step S15). The combustion state of the concentrate can be understood from the classified combustion patterns.

[0051] The combustion determination device 40 according to this embodiment can be realized, for example, by cooperation between hardware constituting a general computer (information processing device) described below and a program (software) executed by the computer. For example, the computer can execute a predetermined program to realize each processing unit, such as the acquisition unit 41, extraction unit 42, learning unit 43A, calculation unit 44A, and classification unit 45. Furthermore, the image storage unit 47 can be realized using a storage device included in the computer.

[0052] As described above, the combustion determination device 40 according to this embodiment includes an acquisition unit 41, an extraction unit 42, a calculation unit 44A, and a classification unit 45, and the classification unit 45 classifies the data into combustion patterns based on the probabilities calculated by the calculation unit 44A. As a result, the combustion determination device 40 can obtain the combustion pattern of the concentrate from the acquired target image data using the classification unit 45, thereby understanding the combustion state of the concentrate. Therefore, the combustion determination device 40 can control the amounts of reactive gases, auxiliary agents, etc. introduced when smelting the concentrate in the flash smelting furnace 1, and can control the combustion state of the concentrate in the flash smelting furnace 1.

[0053] The combustion determination device 40 can include a learning unit 43A. The learning unit 43A can improve the accuracy of calculating the occurrence probability of the concentrate combustion pattern and shorten the time required for the calculation by learning the combustion pattern in advance using a set of target image data acquired by the acquisition unit 41. Therefore, the combustion determination device 40 can obtain the concentrate combustion pattern more accurately and in a shorter time using the classification unit 45, and can therefore grasp the combustion state of the concentrate more accurately and in a shorter time.

[0054] The combustion determination device 40 can use the learning unit 43A to learn combustion patterns using deep learning with a neural network. This allows the learning unit 43A to further increase the accuracy of calculating the occurrence probability of the concentrate combustion pattern and further shorten the time required for the calculation. Therefore, the classification unit 45 of the combustion determination device 40 can obtain the concentrate combustion pattern with even greater accuracy in a shorter time, making it possible to grasp the concentrate combustion state with even greater accuracy and in a shorter time.

[0055] The combustion determination device 40 can use color information for the pixel values ​​extracted by the extraction unit 42. This allows the extraction unit 42 to easily extract the pixel values ​​of the image included in a set of target image data. By using the pixel values ​​extracted by the calculation unit 44A, the combustion determination device 40 can easily calculate the probability representing the similarity with the combustion pattern, making it possible to easily grasp the combustion state of the concentrate.

[0056] Therefore, by including the combustion determining device 40, the control system 100A can control the combustion state in the flash smelting furnace 1 appropriately with high accuracy.

[0057] In this embodiment, the combustion determining device 40 does not necessarily have to include the learning unit 43A.

[0058] In this embodiment, the user may re-specify a classification different from the classification result from the classification unit 45, and the learning unit 43A may again perform learning using the time information of the target image.

[0059] <Second embodiment> A composition identification device according to a second embodiment will be described. The combustion determination device 40 according to the first embodiment classifies a combustion pattern based on a probability representing the degree of similarity with a combustion pattern that represents the combustion state of the concentrate. However, the composition identification device according to this embodiment identifies the composition ratio of the concentrate based on a probability representing the degree of similarity with the composition ratio of the concentrate. The composition identification device according to this embodiment will be described for a case in which the composition ratio of the concentrate is acquired in advance when used for learning, and learning is performed using the pixel values ​​obtained from the captured image and the composition ratio of the concentrate as teacher labels. Note that in this embodiment, a description of the content common to the first embodiment will be omitted, and composition identification that differs from the first embodiment will be described.

[0060] Fig. 6 is a diagram showing the configuration of a control system including a composition identification device according to the second embodiment, and the functional configuration of the composition identification device. As shown in Fig. 6, the control system 100B includes a composition identification device 50. Instead of the calculation unit 44A and classification unit 45 of the combustion determination device 40 according to the first embodiment, the composition identification device 50 includes a calculation unit 44B that calculates the probability of a relationship between time information of a target image capturing the combustion of concentrate and the composition ratio of the concentrate, and an identification unit 48 that identifies the composition of the concentrate from the calculation result.

[0061] The learning unit 43B may learn the composition ratio of the concentrate in advance using a set of target image data acquired by the acquisition unit 41. The learning method is similar to that of the learning unit 43A of the first embodiment described above, except that the content learned in advance using the set of target image data acquired by the acquisition unit 41 is changed from the combustion pattern to the composition ratio of the concentrate, and therefore details will be omitted.

[0062] The calculation unit 44B calculates the occurrence probability of the composition ratio of the concentrate from the pixel value and the time information of the target image containing the pixel value. The occurrence probability of the composition ratio of the concentrate is calculated using a model previously learned by the learning unit 43B.

[0063] Table 1 shows the prediction error (unit: %) when the composition ratio of the copper concentrates in Figures 3(a) and 3(b) is specified. Note that concentrate A in Table 1 represents the copper concentrates in Figures 3(a) and 4(a), and concentrate B represents the copper concentrates in Figures 3(b) and 4(b).

[0064] [Table 1]

[0065] Here, the prediction error is the difference, expressed as a percentage, between a standardized predicted value (for example, a value predicted using a learning model for the proportion of Cu in a concentrate) and a standardized teacher label value (a value obtained by actually quantitatively analyzing the concentrate). Note that standardization can be calculated using the following formulas (1) and (2). In the following formulas (1) and (2), z i is the prediction error, and x i is a predicted value or a teacher label value, μ is the average value of each of the predicted values ​​or teacher label values, σ is the standard deviation of each of the predicted values ​​or teacher label values, and N is an integer of 1 or more.

[0066]

number

[0067] The smaller the prediction error (%), the better. For example, in the case of concentrate A, which had a good combustion state, the prediction accuracy for Cu and SiO2 was good. Note that these specific composition targets can be changed by the user as appropriate if they can be obtained as teaching labels during learning.

[0068] The identifying unit 48 identifies the composition of the target concentrate using the results of the calculation unit 44 B. The identifying unit 48 sends the identified classification results to the control unit 46.

[0069] In this way, the composition identifying device 50 according to this embodiment can obtain the composition ratio of the target object from the target image, and can therefore grasp changes in the composition ratio in the concentrate. Therefore, the composition identifying device 50 can control combustion during smelting, and the addition of reaction gases and auxiliary agents, etc.

[0070] Like the combustion determining device 40, the composition identifying device 50 can include a learning unit 43B. Therefore, in the composition identification device 50, the learning unit 43B can increase the accuracy of identifying the composition ratio of the concentrate and shorten the time required for identification by learning the composition ratio of the concentrate in advance using a set of target image data acquired by the acquisition unit 41. Therefore, the combustion determination device 40 can obtain the composition ratio of the concentrate more accurately and in a shorter time using the identification unit 48, and therefore can grasp changes in the composition ratio in the concentrate more accurately and in a shorter time.

[0071] As with the combustion determination device 40, the composition identification device 50 can use the learning unit 43B to learn the composition ratio of the concentrate using deep learning with a neural network. This allows the learning unit 43B to further increase the accuracy of identifying the composition ratio of the concentrate and further shorten the time required for identification. Therefore, the combustion determination device 40 can obtain the composition ratio of the concentrate with even greater accuracy in a shorter time using the identification unit 48, making it possible to grasp changes in the composition ratio in the concentrate with even greater accuracy and in a shorter time.

[0072] Therefore, by including the composition specifying device 50, the control system 100B can control the combustion state in the flash smelting furnace 1 appropriately with high accuracy.

[0073] In this embodiment, the user may re-specify a classification different from the classification result from the specification unit 48, and the learning unit 43B may again perform learning using the time information of the target image.

[0074] <Hardware configuration example> The combustion determination device 40 and the composition identification device 50 according to this embodiment are implemented by an information processing device. The hardware configuration of the information processing device will be described below with reference to FIG. 7. FIG. 7 is a block diagram showing an example of the hardware configuration of the information processing device. As shown in FIG. 7, the information processing device 200 has a hardware configuration as a general computer, including a processor 210 such as a CPU (Central Processing Unit), memories such as a ROM (Read Only Memory) 220 and a RAM (Random Access Memory) 230, a storage device such as an HDD (Hard Disk Drive) 240, an output unit 250 related to a display device such as a presentation unit that presents results to a user, an input unit 260 related to an input device such as a camera 20 or a reception unit that accepts instructions from a user, an I / F unit 270 for connecting devices such as the output unit 250 and the input unit 260, and a bus connecting these units.

[0075] In this case, the program is provided by being recorded on, for example, a magnetic disk, an optical disk, a semiconductor memory, or a similar recording medium, and is stored in a storage device or the like. The recording medium on which the program is recorded may have any storage format as long as it is a computer-readable recording medium. The program may be configured to be pre-installed on the computer, or may be configured to be distributed via a network and then installed on the computer as needed.

[0076] The program executed by the computer has a modular configuration including the processing units of the combustion determination device 40 and the composition identification device 50 described above. The processor 210 reads and executes this program as appropriate, thereby generating the processing units described above in a memory such as the RAM 230.

[0077] The combustion determination device 40 or the composition identification device 50 according to this embodiment may be configured as a system in which a plurality of computers are connected to each other so as to be able to communicate with each other, and the above-described processing units may be distributed among the plurality of computers to be realized. Alternatively, the combustion determination device 40 or the composition identification device 50 may be a virtual machine operating on a cloud system.

[0078] As described above, several embodiments have been described, but these embodiments are presented as examples and are not intended to limit the scope of the invention. Each of the above embodiments can be implemented in various other forms, and various combinations, omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, and are also included in the inventions and their equivalents as set forth in the claims. [Explanation of symbols]

[0079] 100A, 100B control system 10 Feeding device 20 Camera (imaging device) 30 Combustion equipment 40 Combustion determination device 41 Acquisition Department 42 Extraction part 43A, 43B Learning Department 44A, 44B Calculation section 45 Classification Department 46 Control Unit 47 Image storage unit 48 Specific part 50 Composition identification device

Claims

1. an acquisition unit that acquires, as a set of target image data, each of a plurality of images captured during the process of heating and burning a concentrate containing recovered metals and time information for each of the plurality of images; an extraction unit that extracts pixel values ​​of the image included in the target image data; a calculation unit that calculates a probability representing a similarity between the pixel value and the time information and the composition ratio of the concentrate based on the extracted pixel value and time information included in the target image data; an identification unit that identifies a composition ratio of the concentrate based on the probability; a learning unit that learns in advance the composition ratio of the concentrate using the set of target image data; A composition identification device comprising:

2. A composition identification device as described in Claim 1, wherein the learning unit learns the composition ratio of the concentrate using deep learning with a neural network.

3. A composition identification device as described in claim 1 or 2, in which the pixel values ​​use color information.

4. a supply device for supplying the concentrate containing recovered metals to the combustion furnace; a combustion device that combusts the concentrate in the combustion furnace; an imaging device that images the inside of the combustion furnace through an observation window of the combustion furnace; The composition specifying device according to any one of claims 1 to 3; Equipped with a control system for controlling the combustion device based on the composition ratio of the concentrate identified by the identification unit;

Images