Method for calculating the molding ratio of a molded body

Abstract

This invention provides a method for calculating the molding rate of a molded body with high accuracy. [Solution] The method includes: a distance image acquisition step ST1 in which a molded body M is placed on a belt conveyor BC immediately after molding and an image is taken of it to obtain a first distance image showing the distance from a reference position to the surface of the molded body and the surface of the belt conveyor; a difference image generation step ST3 in which a difference image is generated between the first distance image and a second distance image showing the distance from a reference position to the surface of the belt conveyor, which is obtained in advance by taking an image of the belt conveyor when the molded body is not placed on it; a feature extraction step ST4 in which a histogram showing the relationship between the pixel value and the number of pixels of the pixels constituting the difference image is calculated and feature quantities are extracted from the histogram; and a molding rate estimation step ST6 in which the molding rate of the molded body is estimated based on a calibration curve showing the relationship between the feature quantities of the histogram created in advance and the actual molding rate, and the feature quantities extracted in the feature extraction step.

Description

Technical Field

[0001] The present invention relates to a method for accurately calculating the molding rate of a molded body in which finished particles obtained by molding powder particles into a certain shape, such as molded charcoal used as a raw material for coke, and powder particles remaining without being molded as finished particles are mixed.

Background Art

[0002] In blast furnace operation, in order to ensure the air permeability in the furnace, the coke, which is a blast furnace raw material, is required to have a required strength. For this reason, in the manufacturing process of coke, a method of increasing the bulk density and enhancing the strength of coke by blending molded charcoal previously molded into a certain shape is known. Molded charcoal is manufactured by adding a binder such as tar to pulverized coal obtained by pulverizing raw coal, kneading them, and molding them into lumps of about 20 to 70 cc per piece.

[0003] The manufactured molded charcoal is transported to the coke oven while being placed on a belt conveyor. However, if the molded charcoal is not sufficiently molded and contains many particles with a small particle size (an index indicating the degree of the particle diameter, which is the particle size), it is known that the strength of the coke decreases. For this reason, it is desirable to continuously measure the particle size distribution of the particles constituting the molded charcoal immediately after molding and maintain the quality by improving the manufacturing equipment and manufacturing conditions of the molded charcoal. Specifically, from the measured particle size distribution, the mass ratio of particles larger than a predetermined particle size (for example, a particle size of 10 mm) in the entire particles constituting the molded charcoal is calculated as the molding rate, which is an index indicating the degree of molding. It is desirable to improve the manufacturing equipment and manufacturing conditions of the molded charcoal according to this molding rate.

[0004] The particle size of molded charcoal is generally measured by taking samples from a conveyor belt at regular time intervals and then sieving them about three times a day. Therefore, even if the particle size distribution of molded charcoal fluctuates rapidly due to variations in the quality of the raw materials or malfunctions in the manufacturing equipment, intermittent measurements using sieving after sampling result in coarse time intervals, making it impossible to accurately capture the temporal fluctuations in the particle size distribution.

[0005] As a method for continuously measuring particle size in a non-contact manner, for example, the method described in Non-Patent Document 1 has been proposed. The method described in Non-Patent Document 1 is a measurement method that uses a light-section type 3D camera in which a laser light source that emits linear laser light and an area scan camera are integrated. In the method described in Non-Patent Literature 1, a distance detection means such as a rotary encoder in contact with the conveyor belt is used to measure the position of the upper edge of the cross-section of particles accumulated on the conveyor belt using a 3D camera each time the conveyor belt moves a certain distance. This generates a distance image (sometimes called a 3D image or depth image) in which the pixel value of each pixel indicates the distance from a reference position (for example, the distance from the 3D camera). Near the boundaries of the stacked particles, the irradiated laser light is interrupted and the image becomes darker, and the difference in height of the unevenness of the particles becomes larger. As a result, in the distance image, the pixel values ​​of the pixel regions corresponding to the particle boundaries tend to be different from the pixel values ​​of other pixel regions. In the method described in Non-Patent Literature 1, this characteristic is used to determine the boundaries of the particles, identify each particle, and calculate the particle size of each particle. Among the stacked particles, the dimensions of particles that have parts hidden by other particles will be smaller than their actual dimensions. Therefore, in the method described in Non-Patent Document 1, surface particles (hereinafter referred to as "surface particles") are preferentially extracted using height information of each particle (height from the bottom of the conveyor belt) that can be calculated by a 3D camera, and the diameter of the minor axis when each surface particle in the depth image is considered as an ellipse is used as the particle size.

[0006] As a method for continuously measuring particle size without contact, methods using a 3D camera, similar to the method described in Non-Patent Document 1, have also been proposed, as described in Non-Patent Document 2 and Patent Document 1. Non-patent document 2 and patent document 1 describe in detail edge detection methods for identifying individual particles and image processing methods for recognizing surface particles. In particular, patent document 1 also describes a method for speeding up the measurement process.

[0007] In the method described in Non-Patent Literature 2, image processing is applied to the depth image obtained of the deposited particles to extract surface particles in the surface layer with little overlap, and the particle size of each surface particle is sorted into particle size classifications (classifications determined by particle size) determined by the mesh size of the sieve, and the number distribution of surface particles (surface number distribution), which is the relationship between the particle size classification and the number of surface particles in each particle size classification, is calculated. Furthermore, in the method described in Non-Patent Literature 2, a surface stochastic model, which is a model that represents the degree of appearance and visibility on the surface according to particle size, that is, a model that estimates the number distribution of the entire deposited particles (overall number distribution) from the number distribution of surface particles, is used to estimate the number distribution of the entire deposited particles, including not only surface particles but also hidden particles. In addition, in the method described in Non-Patent Literature 2, the distribution of the total mass ratio of the entire deposited particles (overall mass distribution) is estimated using the volume ratio (or mass ratio) for each particle size classification.

[0008] Here, the present inventors propose a particle size distribution measurement method described in Patent Document 2. In Patent Document 2, the present inventors applied the method described in Non-Patent Document 2 to a mixed particle size distribution deposit in which particles belonging to multiple particle size categories (coke particles) are blended in a predetermined mass ratio and deposited, and conducted a verification test to see whether the overall mass distribution could be estimated with good accuracy. As a result of this verification test, it was found that the particle size of surface particles changes depending on the orientation of the particles, and the particle size distribution is wider than that of the particle size categories. Therefore, if the particle size of surface particles measured with a 3D camera is sorted into particle size categories, the number of surface particles for each particle size category is tallied, and the method described in Non-Patent Document 2 is applied, the particle size distribution (overall mass distribution) becomes blurred and does not accurately match the results measured by sieving. Therefore, in the particle size distribution measurement method described in Patent Document 2, for each of the multiple particle size categories to which the particles constituting the mixed particle size deposit belong, single particle size samples (samples consisting only of particles with the same particle size category) are prepared, and the particle size of the surface particles of each single particle size sample is measured using a 3D camera or the like to calculate a first particle size distribution that shows the relationship between the particle size and number of surface particles of the single particle size sample. Furthermore, for the mixed particle size deposit, which is composed of particles belonging to the multiple particle size categories, the particle size of the surface particles is measured using a 3D camera or the like to calculate a second particle size distribution that shows the relationship between the particle size and number of surface particles of the mixed particle size deposit.Then, the calculated second particle size distribution is approximated by a linear sum of the calculated first particle size distributions, and each coefficient of this linear sum is considered to be the ratio of the number of particles of different particle size categories in the surface of the mixed particle size deposit, assuming that the mixed particle size deposit is composed of a combination of single particle size samples from multiple particle size categories.Thereafter, the overall mass distribution of the mixed particle size deposit is calculated using these coefficients in the same procedure as described in Non-Patent Document 2. According to the particle size distribution measurement method described in Patent Document 2, it is possible to calculate a highly accurate particle size distribution (overall mass distribution) that closely matches the results measured using a sieve.

[0009] Furthermore, the present inventors have proposed a method for calculating the collapse rate described in Patent Document 3. When a mixed particle size deposit is molded coal, if the finished particles of the molded coal collapse during the conveying process on a belt conveyor, the number of particles with a particle size smaller than the finished particles that do not collapse increases. For example, the mass ratio of particles with a particle size of 10 mm or less to the total number of particles constituting the molded coal can be defined as the collapse rate. For this reason, in order to calculate the collapse rate using the particle size distribution measurement method described in Patent Document 2, it is conceivable to prepare a single particle size sample consisting only of particles with a particle size of 10 mm or less, measure the particle size of the surface particles of this single particle size sample using a 3D camera or the like, and calculate the first particle size distribution (relationship between particle size and number of particles on the surface of the single particle size sample). However, particles with a particle size of 10 mm or less include particles smaller than the resolution of the 3D camera, and even if the particle size is larger than the resolution, the irregularities at the particle boundaries are small, making it difficult to identify the particles and making it difficult to calculate an accurate first particle size distribution. Therefore, it is not practical to apply the particle size distribution measurement method described in Patent Document 2 to a single-particle size sample consisting only of particles with a particle size of 10 mm or less, which serves as the basis for calculating the disintegration rate. Therefore, in the method for calculating the collapse rate described in Patent Document 3, a finished sample is prepared as a reference sample, which consists only of finished particles molded into a certain shape (for example, 30 mm < particle size ≤ 35 mm), and an intermediate particle size sample consists only of particles belonging to an intermediate particle size category (for example, 10 mm < particle size ≤ 30 mm), which is a particle size category larger than a predetermined particle size, among the particles generated when the finished particles collapse. Then, similar to the particle size distribution measurement method described in Patent Document 2, the first particle size distribution is calculated for the reference sample, and the second particle size distribution is calculated for the mixed particle size deposit (molded charcoal) which is a mixture of finished particles and particles generated when the finished particles collapse. Subsequently, the overall mass distribution of the mixed particle size deposit is calculated using the same procedure as the particle size distribution measurement method described in Patent Document 2. Then, the collapse rate of the mixed particle size deposit is estimated based on a calibration curve that shows the relationship between the total mass ratio of particles in the intermediate particle size category and the actual collapse rate, and the total mass ratio of particles in the intermediate particle size category in the overall mass distribution of the mixed particle size deposit.

[0010] If we define the collapse rate of molded coal calculated using the collapse rate calculation method described in Patent Document 3 as the mass ratio of particles with a particle size of 10 mm or less to the total number of particles constituting the molded coal, and define the molding rate of molded coal as the mass ratio of particles with a particle size larger than 10 mm to the total number of particles constituting the molded coal, then the relationship molding rate [%] = 100 - collapse rate [%] holds true. For this reason, it is conceivable that the molding rate of molded coal immediately after molding can also be calculated using the method described in Patent Document 3, similar to the collapse rate of molded coal in which the finished particles have collapsed during the conveying process on a belt conveyor. Figure 1 shows an example of the cumulative mass distribution measured by sieving molded charcoal. The horizontal axis of Figure 1 represents particle size, and the vertical axis represents the cumulative mass percentage obtained by accumulating the total mass percentage of particles belonging to each particle size category. Figure 1 shows the cumulative mass distribution measured for molded charcoal (No. 1, No. 2) in which the finished particles have broken down during the conveying process on a belt, and the cumulative mass distribution measured for molded charcoal (No. 3, No. 4, No. 5) immediately after molding. For example, in Figure 1, when the value on the horizontal axis is 10 mm, the value on the vertical axis represents the total mass percentage of particles with a particle size of 10 mm or less, and when the value on the horizontal axis is 15 mm, the value on the vertical axis represents the total mass percentage of particles with a particle size of 15 mm or less. As can be seen from Figure 1, for molded coal No. 1 and No. 2, where collapse has occurred, the cumulative mass proportion increases in the intermediate particle size category of 10 mm < particle size ≤ 30 mm, whereas for molded coal No. 3 to No. 5 immediately after molding, the cumulative mass proportion in the intermediate particle size category remains almost constant. In other words, it can be seen that molded coal No. 3 to No. 5 immediately after molding contains almost no particles belonging to the intermediate particle size category. For this reason, it is difficult to accurately calculate the molding rate using the method described in Patent Document 3, which estimates the collapse rate based on the total mass proportion of particles in the intermediate particle size category in the overall mass distribution. [Prior art documents] [Patent Documents]

[0011] [Patent Document 1] Japanese Patent Publication No. 2019-174155 [Patent Document 2] Japanese Patent Publication No. 2022-172620 [Patent Document 3] Japanese Patent Publication No. 2024-035714 [Non-patent literature]

[0012] [Non-Patent Document 1] MJ Thurley, "Automated, On-line, Calibration-Free, Particle Size Measurement using 3D Profile Data", Measurement and Analysis of Blast Fragmentation: Workshop Hosted by FRAGBLAST 10 - The 10th International Symposium on Rock Fragmentation by Blasting, 2013, pp.23-32 [Non-Patent Document 2] MJ Thurley, "Three Dimensional Data Analysis for the Separation and Sizing of Rock Piles in Mining", Ph.D. Thesis, Monash University, December 2002, chapter 4, pp.27-60 [Overview of the project] [Problems that the invention aims to solve]

[0013] The present invention aims to provide a method for calculating the molding ratio of a molded body in which finished particles obtained by molding powder particles into a certain shape, such as molded charcoal used as a raw material for coke, and powder particles that remain without being molded into finished particles are mixed. [Means for solving the problem]

[0014] To solve the aforementioned problems, the inventors conducted diligent research. Specifically, the inventors obtained a first distance image by imaging a molded body, such as molded charcoal, which is a mixture of finished particles obtained by molding powder particles into a certain shape and powder particles that remain unmolded as finished particles, using a 3D camera or the like while the molded body is placed on a belt conveyor immediately after molding. They then generated a difference image between this first distance image and a second distance image obtained by imaging the belt conveyor when the molded body is not placed on it. Since the cross-section of a typical belt conveyor for transporting molded bodies such as molded charcoal is trapezoidal, even if multiple molded bodies of the same particle size are placed on the belt conveyor in a single layer, height differences will occur depending on the cross-sectional shape of the belt conveyor on which each molded body is placed. Therefore, the pixel values ​​of the pixels corresponding to each molded body in the first distance image will be different. However, in the difference image between the first distance image and the second distance image obtained by imaging the belt conveyor without the molded body on it, the influence of the cross-sectional shape of the belt conveyor is eliminated. Therefore, this difference image becomes a distance image (a distance image projected onto a plane) that is similar to the state when the molded body placed on the belt conveyor is placed on a horizontal plane. The inventors then conducted thorough studies using this difference image. As a result, we found a good correlation between the histogram features, which show the relationship between pixel values ​​and the number of pixels constituting the difference image, and the actual molding rate of the molded body (the molding rate calculated by the particle size measured using a sieve). Therefore, for example, by extracting the histogram features from the difference images of multiple molded bodies with different molding conditions and measuring the actual molding rate using a sieve, we found that if a calibration curve showing the relationship between the histogram features and the actual molding rate is created in advance, the molding rate of the target molded body can be calculated (estimated) with high accuracy based on this calibration curve and the histogram features extracted from the difference image of the molded body for which the molding rate is to be estimated.

[0015] This invention was completed based on the findings of the inventors described above. In other words, to solve the above problems, the present invention provides a method for calculating the molding rate of a molded body, comprising: a distance image acquisition step of acquiring a first distance image showing the distance from a reference position to the surface of the molded body and the surface of the belt conveyor by imaging a molded body, which is a mixture of finished particles obtained by molding powder particles into a certain shape and the powder particles that remain unmolded as finished particles, while the molded body is placed on a belt conveyor immediately after molding; a difference image generation step of generating a difference image between the first distance image and a second distance image showing the distance from the reference position to the surface of the belt conveyor, which is acquired in advance by imaging the belt conveyor when the molded body is not placed on it; a feature extraction step of calculating a histogram showing the relationship between the pixel value and the number of pixels of the pixels constituting the difference image and extracting feature quantities from the histogram; and a molding rate estimation step of estimating the molding rate of the molded body based on a calibration curve showing the relationship between the feature quantities of the histogram created in advance and the actual molding rate, and the feature quantities of the histogram extracted in the feature extraction step.

[0016] In the present invention, "a state in which the molded body is placed on a belt conveyor immediately after molding" means a state in which the molded body has not substantially collapsed immediately after being placed on a belt conveyor, or even after being transported by a belt conveyor, due to a short transport distance. In the present invention, "distance image" (first distance image and second distance image) means an image in which the pixel value of each pixel indicates the distance from a reference position (for example, the distance from the distance image acquisition means). The first distance image is an image showing the distance from the reference position to the surface of the molded body and the surface of the conveyor belt, and the second distance image is an image showing the distance from the reference position to the surface of the conveyor belt. The distance image acquisition means for acquiring the distance image is not particularly limited as long as it is a means that can acquire the distance to the object, but for example, a light-section type 3D camera in which a laser light source that emits linear laser light and an area scan camera are integrated can be cited. The reference position can be set to any position, for example, the position of the distance image acquisition means can be set as the reference position.

[0017] According to the present invention, by executing a distance image acquisition step, a difference image generation step, and a feature quantity extraction step, for a molded body to be estimated for the molding rate, feature quantities of a histogram of a difference image between a first distance image and a second distance image are extracted. According to the findings of the inventors described above, there is a good correlation between the feature quantities of the histogram of the difference image and the actual molding rate of the molded body. Therefore, in the molding rate estimation step, using a calibration curve showing the relationship between the feature quantities of the histogram created in advance and the actual molding rate, based on this calibration curve and the feature quantities of the histogram of the difference image of the molded body to be estimated, it is possible to accurately estimate the molding rate of the molded body. Note that the calibration curve used in the molding rate estimation step can be created in advance by preparing a plurality of molded bodies with different molding states separately from the molded body to be estimated for the molding rate, extracting the feature quantities of the histogram of the difference image for each of the plurality of molded bodies, and measuring the actual molding rate using a sieve.

[0018] According to the findings of the inventors, when the histogram of the difference image shows the relationship between the pixel value and the number of pixels, with the pixel value on the horizontal axis and the number of pixels on the vertical axis, if a value proportional or inversely proportional to the difference between the number of pixels corresponding to the positive peak position of the histogram and the number of pixels corresponding to the negative peak position of the histogram is used as the feature quantity of the histogram, a high correlation with the molding rate of the molded body can be obtained. Therefore, in the present invention, preferably, the histogram shows the relationship between the pixel value and the number of pixels, with the pixel value on the horizontal axis and the number of pixels on the vertical axis, and the feature quantity of the histogram is a value proportional or inversely proportional to the difference between the number of pixels corresponding to the positive peak position of the histogram and the number of pixels corresponding to the negative peak position of the histogram.

Effect of the Invention

[0019] According to the present invention, it is possible to accurately calculate the molding rate of a molded body in which finished particles obtained by molding powder particles into a certain shape, such as molded coke used as a raw material for coke, and powder particles remaining without being molded as finished particles, are mixed.

Brief Description of Drawings

[0020] [Figure 1] The figure which shows an example of the cumulative mass distribution measured by sieving the molded coal. [Figure 2] The figure which shows typically the apparatus configuration for carrying out the method for calculating the molding rate of the molded object which concerns on one Embodiment of this invention. [Figure 3] The flowchart which shows the process which the method for calculating the molding rate of the molded object which concerns on one Embodiment of this invention has. [Figure 4] The figure which shows an example of the 1st distance image, the 2nd distance image, and the difference image. [Figure 5] The figure which shows an example of the histogram of the difference image. [Figure 6] Examples of the histograms of the 1st distance image and the difference image acquired for each molded object M with different molding states are shown. [Figure 7] The figure which shows an example of the result of continuously extracting C / A which is a feature quantity of the histogram in the process ST4 shown in FIG. 3 during the conveyance process of the molded object M. [Figure 8] The figure which shows an example of the calibration curve created by the process ST5 shown in FIG. 3.

Mode for Carrying Out the Invention

[0021] Hereinafter, while appropriately referring to the attached drawings, one embodiment of the present invention will be described by taking as an example the case where the molded object is molded coal used as a raw material for coke. FIG. 2 is a diagram schematically showing an apparatus configuration for executing a method for calculating the molding rate of a molded object according to the present embodiment. FIG. 3 is a flowchart showing the steps of the method for calculating the molding rate of a molded object according to the present embodiment. As shown in Figure 2, the molding machine 1 produces finished particles F by molding powdered coal, which is obtained by crushing raw coal and adding a binder such as tar, into a certain shape. The powdered particles P are then mixed in a kneader (not shown) and molded into a specific shape. The finished particles F produced by the molding machine 1, along with the powdered particles P that remain unmolded as finished particles F, are discharged onto a belt conveyor BC. The molded body (molded coal) M, which is a mixture of finished particles F and powdered particles P, is then transported by multiple belt conveyors BC in the direction of the thick arrows shown in Figure 2 and charged into a coke oven (not shown). The molding rate calculation method according to this embodiment is performed using a 3D camera (a 3D camera using the light sectioning method with a linear laser beam L) 2 as a distance image acquisition means positioned above the belt conveyor BC, and a computing device 3 connected to the 3D camera 2. In order to capture images of the molded body M as it is placed on the belt conveyor BC immediately after molding, the 3D camera 2 is positioned above the first belt conveyor BC (the belt conveyor BC located directly below the molding machine 1) on the downstream side in the transport direction.

[0022] In this embodiment, a Sick Ranger-E50 (area scan camera pixels: 1536 x 512, measurement speed: 35,000 sections / second) was used as the 3D camera 2. The height from the surface of the belt conveyor BC to the 3D camera 2 was adjusted so that the 500 mm width of the belt conveyor BC on which the molded body M formed by the molding machine 1 was placed was within the field of view of the area scan camera. The conveying speed of the belt conveyor BC was set to a constant value, and it was determined by the elapsed time that a predetermined conveying distance had been covered. A distance image (a first distance image acquired in step ST1 described later, and a second distance image acquired in step ST2 described later) with a length corresponding to this conveying distance was then acquired in the conveying direction.

[0023] As shown in Figure 3, the molding rate calculation method according to this embodiment comprises steps ST1 to ST6. Each step will be described in order below.

[0024] <Process ST1> In step ST1 (corresponding to the distance image acquisition step of the present invention), as shown in Figure 2, a 3D camera 2 is used to image the molded body M to be used for estimation of the molding rate, and a first distance image is acquired that shows the distance from a reference position (for example, the position of the 3D camera 2) to the surface of the molded body M and the surface of the belt conveyor BC. The laser beam L emitted from the 3D camera 2 extends linearly in a direction perpendicular to the conveying direction of the belt conveyor BC (a direction perpendicular to the plane of the paper in Figure 2), and the distance of the portion irradiated by the laser beam L is measured sequentially in the manner of light sectioning. Then, as the molded body M is conveyed by the belt conveyor BC in a direction perpendicular to the direction in which the laser beam L extends, a first distance image is acquired, which is composed of pixels arranged in two dimensions, and the pixel value of each pixel shows the distance from the reference position. The acquired first distance image is input to a computing device 3 connected to the 3D camera 2.

[0025] <Process ST2> In process ST2, a 3D camera 2 is used to image the conveyor belt BC when no molded body M is placed on it, and a second distance image is obtained that shows the distance from a reference position to the surface of the conveyor belt BC. The second distance image, like the first distance image, is composed of pixels arranged in two dimensions, and the pixel value of each pixel indicates the distance from the reference position. The acquired second distance image is input to the computing device 3 connected to the 3D camera 2. Process ST2 can be executed at any time, as long as it is performed before executing process ST3, which will be described later. Furthermore, unlike process ST1, which is executed for each molded body M for which the molding rate is to be estimated, process ST2 only needs to be executed at least once. In other words, the same second distance image obtained by executing it once can be stored and reused repeatedly in process ST3, which will be described later.

[0026] <Process ST3> In step ST3 (corresponding to the difference image generation step of the present invention), the arithmetic unit 3 generates a difference image between the first distance image and the second distance image acquired by the 3D camera 2. That is, the arithmetic unit 3 subtracts the pixel values ​​of the pixels constituting the second distance image (pixels located at positions corresponding to the pixels constituting the first distance image) from the pixel values ​​of the pixels constituting the first distance image and sets the result as the pixel values ​​of the pixels constituting the difference image. If the result of subtracting the pixel values ​​of the pixels constituting the second distance image from the pixel values ​​of the pixels constituting the first distance image is a negative value, the arithmetic unit 3 performs a saturation operation to round the pixel values ​​of the pixels that have become negative to 0. Figure 4 shows an example of the first distance image, second distance image, and difference image. The upper figures in Figures 4(a) to 4(c) show the first distance image, second distance image, and difference image, respectively. The lower figures in Figures 4(a) to 4(c) schematically show the state of the molded body M and conveyor belt BC when the images shown in the upper figures were acquired and generated (viewed from the conveyor belt BC's conveying direction). Note that the difference image shown in Figure 4(c) is dark as is, so for convenience, the brightness and contrast have been adjusted to make it easier to see. However, in step ST4 described later, the histogram will be calculated using the difference image before brightness and contrast adjustment. The first distance image shown in Figure 4(a) and the second distance image shown in Figure 4(b) both have 256 pixel values ​​for each pixel. The smaller the value (darker), the longer the distance from the 3D camera 2 (i.e., the lower the pixel is located), and the larger the value (brighter the pixel is located), the shorter the distance from the 3D camera 2 (i.e., the higher the pixel is located). In the first distance image shown in Figure 4(a), the black pixels (pixels with a pixel value of 0) correspond to the shadowed areas that could not be captured by the 3D camera 2, and represent areas where the distance could not be measured correctly.

[0027] As can be seen from the first distance image shown in Figure 4(a), the molded body M immediately after molding is placed on the conveyor belt BC in an almost single layer. Since the cross-section of a typical conveyor belt BC is trapezoidal, even if multiple molded bodies M of the same particle size are placed on the conveyor belt BC in a single layer, differences in height will occur depending on the cross-sectional shape of the conveyor belt BC on which each molded body M is placed. Therefore, the pixel values ​​of the pixels corresponding to each molded body in the first distance image will be different. However, in the difference image between the first and second distance images shown in Figure 4(c), the influence of the cross-sectional shape of the conveyor belt BC is eliminated, so this difference image becomes a distance image (a distance image projected onto a plane) that is similar to the state in which the molded body M placed on the conveyor belt BC is placed on a horizontal plane H. As mentioned above, since the molded body M is placed on the conveyor belt BC in an almost single-layer state immediately after molding, if we assume that the molded body M is completely molded (i.e., the molded body M is composed only of finished particles F), then when we calculate the histogram of the difference image (a histogram showing the relationship between pixel values ​​and pixel counts, with pixel values ​​on the horizontal axis and pixel counts on the vertical axis), we can expect that the histogram will have a positive peak at the pixel value corresponding to the height of the molded body M.

[0028] <Process ST4> In step ST4 (corresponding to the feature extraction step of the present invention), the computing unit 3 calculates a histogram showing the relationship between the pixel values ​​and the number of pixels that make up the difference image. Figure 5 shows an example of a histogram of a difference image. As shown in Figure 5, in this embodiment, the arithmetic unit 3 calculates a histogram showing the relationship between pixel values ​​and the number of pixels, with the pixel values ​​on the horizontal axis and the number of pixels on the vertical axis. In reality, the histogram shown in Figure 5 has a very large number of pixels where the pixel value is 0 (pixels corresponding to the surface of the conveyor BC, which is the background of the difference image), but including these would make it difficult to grasp the characteristics of the histogram, so they are omitted from the illustration. The same applies to the histogram shown in Figure 6, which will be described later.

[0029] Then, in step ST4, the arithmetic unit 3 extracts features from the histogram. In this embodiment, as features that show a high correlation with the molding rate of the molded body M, values ​​that are proportional to or inversely proportional to the difference (C shown in Figure 5, where C=AB) between the number of pixels corresponding to the positive peak position of the histogram (A shown in Figure 5) and the number of pixels corresponding to the negative peak position (B shown in Figure 5) are extracted. Examples of values ​​that are proportional to or inversely proportional to C include C, C / A, A / C, C / B, B / C, etc. If the discharge rate of the molded body M discharged from the molding machine 1 onto the belt conveyor BC is constant (the amount of molded body M discharged per unit time) and the conveying speed of the belt conveyor BC is constant, then C can be used as a feature. However, in other cases, the number of pixels corresponding to the molded body M in the difference image changes depending on the discharge rate of the molded body M and the conveying speed of the belt conveyor BC, so it is preferable to use normalized values ​​of C, such as C / A and C / B, or their reciprocals, such as A / C and B / C, as feature quantities.

[0030] Figure 6 shows examples of histograms of the first distance image and difference image obtained for each molded body M with different molding conditions. Figure 6(a) shows the histogram of the first distance image and difference image obtained for a molded body M with good molding condition, Figure 6(b) shows the histogram of the first distance image and difference image obtained for a molded body M with poor molding condition, and Figure 6(c) shows the histogram of the first distance image and difference image obtained for a molded body M with very poor molding condition. The left figures in Figures 6(a) to 6(c) show the first distance image, and the right figures in Figures 6(a) to 6(c) show the histogram of the difference image. As can be seen by comparing the histograms in Figure 6(a) and Figure 6(b), when the molding quality deteriorates (in other words, the molding rate decreases), values ​​proportional to C, such as C, C / A, and C / B, decrease. Furthermore, as can be seen from the histogram in Figure 6(c), when the molding quality deteriorates further, both the positive and negative peak positions disappear. In cases where the peak positions disappear in this way, we can decide to set the values ​​proportional to C to 0. Thus, it can be seen that there is a positive correlation between the molding rate of the molded body M and the histogram features when the values ​​proportional to C are set. On the other hand, as can be seen by comparing the histograms in Figure 6(a) and Figure 6(b), when the molding quality deteriorates (in other words, the molding rate decreases), values ​​inversely proportional to C, such as A / C and B / C, increase. Then, as shown in Figure 6(c), if the peak position disappears, we can decide to set a value inversely proportional to C to a predetermined large value. In this way, we can see that there is a negative correlation between the molding rate of the molded body M and the histogram features when the value is inversely proportional to C.

[0031] Figure 7 shows an example of the results obtained by continuously extracting the histogram feature C / A during the transport process of the molded body M in process ST4. Figures 7(a) to 7(c) show the results obtained when the molding state of the molded body M is good, while Figures 7(d) and 7(e) show the results obtained when the molding state of the molded body M is poor. Figures 7(a) to 7(e) also show the results of measuring the actual molding rate of the molded body M using a sieve. Specifically, the molded body M was removed at the timing when it moved from the first belt conveyor BC to the next belt conveyor BC shown in Figure 2 (i.e., the timing when it is considered that the collapse of the molded body M has not substantially occurred), and the actual molding rate (in this embodiment, the mass ratio of particles with a particle size of 10 mm or larger to the total number of particles constituting the molded body M) was measured using a sieve. Figures 7(a) to 7(e) show, with thick arrows, the time when the molded body M, whose molding rate was measured, was thought to be located directly below the 3D camera 2, along with the measured molding rate value.

[0032] Figure 7(a) shows the C / A ratio extracted during the time from the completion of the manufacturing of the molded body M in the previous production batch by molding machine 1 to the start of the manufacturing of the molded body M in the subsequent production batch by molding machine 1. As shown in Figure 7(a), the C / A value decreases during the process of completing the manufacturing of the molded body M in the previous production batch, and since no molded body M is manufactured between the end of the previous production batch and the start of the subsequent production batch, the C / A value becomes a small, almost constant value, and then increases during the process of starting the manufacturing of the molded body M in the subsequent production batch. This time-series change in the C / A value corresponds well to the manufacturing process of the molded body M by molding machine 1. Figures 7(b) and 7(c) show the C / A extraction results in the middle of a manufacturing batch. As shown in Figures 7(b) and 7(c), it can be seen that the variation in C / A values ​​is small when the molding condition is good. Figures 7(d) and 7(e) show the results of extracting C / A in the middle of a manufacturing batch. The results shown in Figure 7(d) and Figure 7(e) are obtained when the molding conditions of molding machine 1 are shifted in opposite directions from the normal conditions that should be set. In the case shown in Figure 7(d), the C / A value is at best about half of the value when the molding condition is good, as shown in Figures 7(a) to 7(c), indicating that the C / A value decreases as the molding condition deteriorates. In the case shown in Figure 7(e), the C / A value can be as large as when the molding condition is good, but it can also be as small as in Figure 7(d), indicating that there is a very large variation in the C / A value. Thus, simply by understanding the time-series changes in the C / A ratio, it is possible to evaluate the quality of the molded product M.

[0033] <Process ST5> In step ST5, a calibration curve is created that shows the relationship between the histogram features of the difference image and the actual molding rate. Specifically, in addition to the molded body M for which the molding rate is to be estimated, multiple molded bodies with different molding states (different molding rates) are prepared. Then, for each of these multiple molded bodies, the aforementioned steps ST1 to ST4 are performed using the 3D camera 2 and the computing device 3 to extract the histogram features (C / A, etc.) of the difference image. Meanwhile, each of the multiple molded bodies is removed, for example, at the timing when it moves from the first belt conveyor BC to the next belt conveyor BC as shown in Figure 2, and the actual molding rate is measured using a sieve. This makes it possible to create a calibration curve that shows the relationship between the histogram features of the difference image and the actual molding rate. The created calibration curve is stored in the computing device 3. Process ST5 can be executed at any time, as long as it is performed before executing process ST6, which will be described later. Furthermore, unlike processes ST1, ST3, and ST4, which are performed for each molded body M for which the molding rate is to be estimated, process ST5 only needs to be executed at least once. In other words, the calibration curve created and stored after executing it once can be repeatedly used in process ST6, which will be described later.

[0034] Figure 8 shows an example of a calibration curve created by process ST5. The calibration curve shown by the dotted line in Figure 8 is the calibration curve created when C / A is used as the feature quantity of the histogram. It was created by applying the least squares method to the data points extracted and measured as described above (points plotted as black circles in Figure 8) and approximating them with a predetermined function y=f(x) (where x is the molding rate and y is the feature quantity). Note that when the calibration curve shown in Figure 8 was created, only a small amount of data was available for the extreme cases of high and low molding rates, so the function shown in Figure 8 is a first-order polynomial. However, if a large amount of data is available for a wide range of molding rates, it is not limited to this and various functions such as second-order or third-order polynomials can be used. As shown in Figure 8, a good correlation can be seen between the histogram feature ratio C / A and the actual molding rate of the molded product.

[0035] <Process ST6> In step ST6 (corresponding to the molding rate estimation step of the present invention), the computing device 3 estimates the molding rate of the molded body based on the stored calibration curve and the feature quantities extracted in step ST4 for the molded body M whose molding rate is to be estimated. For example, as shown in Figure 8, if the feature quantity (C / A) extracted from the molded body M for which the molding rate is to be estimated is y1, the computing device 3 uses the calibration curve y=f(x) to estimate x1 that satisfies y1=f(x1) as the molding rate of the molded body M. To calculate x1 that satisfies y1=f(x1), known numerical calculation methods such as Newton's method can be used as needed.

[0036] According to the molding rate calculation method of this embodiment described above, by executing steps ST1 to ST4, the histogram features of the difference image between the first distance image and the second distance image are extracted for the molded body M whose molding rate is to be estimated. There is a good correlation between the histogram features of the difference image and the actual molding rate of the molded body M, as shown in Figure 8. Therefore, in step ST6, using a calibration curve that shows the relationship between the histogram features and the actual molding rate, which was created in step ST5, it is possible to accurately estimate the molding rate of the molded body M based on this calibration curve and the histogram features of the difference image of the molded body M to be estimated. [Explanation of symbols]

[0037] 1...Molding machine 2. 3D camera (means for acquiring distance images) 3...Arithmetic unit BC... Belt conveyor F...product particles M...Molded body P...Powder particles ST1...Process (Distance image acquisition step) ST2...Process ST3... Process (Difference Image Generation Step) ST4... Process (Feature Extraction Step) ST5...Process ST6...Process (Molding rate estimation step)

Claims

1. A distance image acquisition step involves capturing an image of a molded body, which is a mixture of finished particles obtained by molding powder particles into a certain shape and the powder particles that remain unmolded as finished particles, while the molded body is placed on a belt conveyor immediately after molding, thereby acquiring a first distance image showing the distance from a reference position to the surface of the molded body and the surface of the belt conveyor. A difference image generation step of generating a difference image between the first distance image and a second distance image obtained in advance by imaging the belt conveyor when the molded body is not placed on it, which shows the distance from the reference position to the surface of the belt conveyor, A feature extraction step is performed to calculate a histogram showing the relationship between the pixel values ​​and the number of pixels that make up the difference image, and to extract the features from the histogram. A molding rate estimation step that estimates the molding rate of the molded body based on a calibration curve showing the relationship between the features of the histogram created in advance and the actual molding rate, and the features of the histogram extracted in the feature extraction step, A method for calculating the molding ratio of a molded body, comprising [the specified parameters].

2. The histogram shows the relationship between the pixel value and the number of pixels, with the pixel value on the horizontal axis and the number of pixels on the vertical axis. The method for calculating the molding rate of a molded body according to claim 1, wherein the feature quantity of the histogram is a value that is proportional to or inversely proportional to the difference between the number of pixels corresponding to the positive peak position of the histogram and the number of pixels corresponding to the negative peak position of the histogram.

Images