JP2024140886A - Granular material particle size measuring device and granular material particle size measuring method - Google Patents

Abstract

[Problem] To provide a particle size measuring device and a particle size measuring method for a granular material that can measure the particle size distribution of a granular material online while taking into account granular materials present inside the granular material loaded on a conveyor. [Solution] A particle size measuring device for a granular material that measures the particle size distribution of a granular material loaded on a conveyor and transported includes a line sensor that creates line profile data of the granular material at a plurality of different positions in the transport direction, and an image processing device, the image processing device having a three-dimensional shape creation section that uses the plurality of line profile data created by the line sensor to create three-dimensional shape data of the granular material, and a particle size calculation section that uses the three-dimensional shape data to measure the particle size and number of the granular material and calculates the particle size distribution of the granular material by assuming that the area ratio of the granular material is the weight ratio of the granular material. [Selected Figure] Figure 2

Description

The present invention relates to a particle size measuring device and a particle size measuring method for granular materials that can measure the particle size distribution of granular materials loaded and transported on a conveyor online.

Coal plays an important role as a raw material for manufacturing steel products. Coal is used as a raw material for producing coke in coke ovens. Coke is used as a reducing agent for iron ore. Currently, the caking coal required for coke production is on the verge of being depleted worldwide, and using high-quality caking coal to produce coke leads to a significant increase in costs. In order to produce high-strength coke using low-quality coal, it is necessary to mold the low-quality coal into molded coal before charging it into the coke oven, and then charge the molded coal into the coke oven for carbonization.

When it comes to molded coal, particle size is one of the important factors. The reason is that if the molded coal is destroyed by vibrations and shocks during transportation before being charged into the coke oven and the particle size becomes small, it will not be possible to produce high-strength coke. Even when using coal as is without molding it, it is very important to measure the particle size distribution of the coal in order to understand its condition.

The particle size distribution of granular materials such as molded coal and coal is obtained by sieving them through multiple sieves with different mesh sizes and calculating the frequency of each particle size. In this way, the particle size measurement method using sieving requires a considerable amount of time before the measurement results are obtained. Therefore, if the particle size distribution changes due to a change in the manufacturing conditions of the granular material, a considerable amount of time is required before the change can be detected. In order to quickly detect a change in the particle size distribution of molded coal, it is necessary to measure the particle size of the granular material online. As a technology for measuring the particle size of granular raw materials online, Patent Document 1 discloses a particle size measurement device having a particle size calculation means that measures the distance between each particle and an imaging unit in image data obtained by capturing an image of the granular raw material moving on a transport route, and calculates the particle size of the granular raw material based on the distance. According to Patent Document 1, the particle size measurement device can be used to measure the particle size of the granular raw material moving on a transport route online.

JP 2014-92494 A

In Patent Document 1, the particle size of the granular raw material is measured using image data of the granular raw material. Therefore, the particle size measured by the particle size measuring device disclosed in Patent Document 1 takes into account only the granular raw material present on the surface layer of the granular raw material loaded on the conveying path, and does not take into account the granular raw material present inside. Therefore, the particle size measuring device disclosed in Patent Document 1 has a problem that the accuracy of measuring the particle size of the granular raw material may be low depending on the internal state of the granular raw material loaded on the conveying path. The present invention has been made in consideration of such problems in the conventional technology, and its purpose is to provide a particle size measuring device and a particle size measuring method for granular material that can measure the particle size distribution of granular material online while taking into account the granular material present inside the granular material loaded on the conveyor.

上記（１）、（２）式において、ｎ_ｉは粒径Ｄ_ｉの粒状物の数（個）であり、ｎ_ｓｉは前記数の測定値（個）であり、αは定数であり、Ｄ_ｉは前記粒状物の粒径（ｍｍ）であり、Ｗ_ｉは粒径Ｄ_ｉの粒状物の重量度数である。
［３］前記粒度算出部は、前記３次元形状データが予め定められた条件を満足する場合に、前記３次元形状データを用いて前記粒状物の重量度数を算出し、前記３次元形状データが予め定められた条件を満足しない場合には、前記３次元形状データを用いて前記粒状物の重量度数を算出しない、［１］又は［２］に記載の粒状物の粒度測定装置。
［４］前記粒度算出部は、前記３次元形状データの高さの平均値を前記粒状物の平均粒径で除した正規化平均高さが３～８の範囲内であり、高さの標準偏差を前記粒状物の平均粒径で除した正規化標準偏差が０．５～１．８の範囲内である場合に、前記条件を満足すると判断し、前記正規化平均高さ及び前記正規化標準偏差の少なくとも１つが上記範囲外である場合に前記条件を満足しないと判断する、［３］に記載の粒状物の粒度測定装置。
［５］前記粒度算出部は、前記３次元形状データの高さの搬送方向の標準偏差の幅方向の平均値が、予め定められた閾値よりも大きい場合に前記条件を満足すると判断し、前記高さの搬送方向の標準偏差の幅方向の平均値が前記閾値以下である場合に、前記条件を満足しないと判断する、［３］又は［４］に記載の粒状物の粒度測定装置。
［６］前記粒度算出部は、ラインプロファイルデータを作成している間の前記コンベアの速度の変化量が予め定められた範囲内である場合に前記条件を満足すると判断し、前記コンベアの速度の変化量が予め定められた範囲外である場合に前記条件を満足しないと判断する、［３］から［５］のいずれかに記載の粒状物の粒度測定装置。
［７］前記粒度算出部は、入力が３次元形状データであり、出力が前記条件を満足するか否かを示すデータである学習済みの機械学習モデルに、前記３次元形状作成部によって作成された３次元形状データを入力し、前記学習済の機械学習モデルから前記条件を満足することを示すデータが出力された場合に前記条件を満足すると判断し、前記学習済の機械学習モデルから前記条件を満足しないことを示すデータが出力された場合に前記条件を満足しないと判断する、［３］に記載の粒状物の粒度測定装置。
［８］前記機械学習モデルは、３次元形状データと、前記３次元形状データが前記条件を満足するか否かを示すデータとを１組とする複数のデータセットを教師データとして機械学習されるディープラーニングネットワークである、［７］に記載の粒状物の粒度測定装置。
［９］コンベア上に積載されて搬送される粒状物の粒度分布を測定する粒状物の粒度測定方法であって、ラインセンサを用いて、搬送方向に異なる複数の位置における前記粒状物のラインプロファイルデータを作成するラインプロファイルデータ作成ステップと、複数のラインプロファイルデータを用いて前記粒状物の３次元形状データを作成する３次元形状データ作成ステップと、前記３次元形状データを用いて、前記粒状物の粒径と数を測定するとともに、前記粒状物の面積比率が前記粒状物の重量比率であるとして、前記粒状物の粒度分布を算出する、粒度算出ステップと、を有する、粒状物の粒度測定方法。
［１０］前記粒度算出ステップでは、前記粒径と前記数と下記（１）、（２）式とを用いて前記粒状物の粒度分布を算出する、［９］に記載の粒状物の粒度測定方法。

上記（１）、（２）式において、ｎ_ｉは粒径Ｄ_ｉの粒状物の数（個）であり、ｎ_ｓｉは前記数の測定値（個）であり、αは定数であり、Ｄ_ｉは前記粒状物の粒径（ｍｍ）であり、Ｗ_ｉは粒径Ｄ_ｉの粒状物の重量度数である。
［１１］前記粒度算出ステップでは、前記３次元形状データが予め定められた条件を満足する場合に、前記３次元形状データを用いて前記粒状物の重量度数を算出し、前記３次元形状データが予め定められた条件を満足していない場合には、前記３次元形状データを用いて前記粒状物の重量度数を算出しない、［９］又は［１０］に記載の粒状物の粒度測定方法。
［１２］前記粒度算出ステップでは、前記３次元形状データの高さの平均値を前記粒状物の平均粒径で除した正規化平均高さが３～８の範囲内であり、高さの標準偏差を前記粒状物の平均粒径で除した正規化標準偏差が０．５～１．８の範囲内である場合に、前記条件を満足するとし、前記正規化平均高さ及び前記正規化標準偏差の少なくとも１つが上記範囲外である場合に前記条件を満足しないとする、［１１］に記載の粒状物の粒度測定方法。
［１３］前記粒度算出ステップでは、前記３次元形状データの高さの搬送方向の標準偏差の幅方向の平均値が予め定められた閾値より大きい場合に前記条件を満足するとし、前記高さの搬送方向の標準偏差の幅方向の平均値が前記閾値以下である場合に前記条件を満足しないとする、［１１］又は［１２］に記載の粒状物の粒度測定方法。
［１４］前記粒度算出ステップでは、ラインプロファイルデータを作成している間の前記コンベアの速度の変化量が予め定められた範囲内である場合に前記条件を満足するとし、前記コンベアの速度の変化量が予め定められた範囲外である場合に前記条件を満足しないとする、［１１］から［１３］のいずれかに記載の粒状物の粒度測定方法。 The means for solving the above problems are as follows.
[1] A particle size measuring device for a granular material that measures the particle size distribution of a granular material loaded and transported on a conveyor, the device comprising: a line sensor that creates line profile data of the granular material at a plurality of different positions in a transport direction; and an image processing device, the image processing device having a three-dimensional shape creation unit that creates three-dimensional shape data of the granular material using the plurality of line profile data created by the line sensor; and a particle size calculation unit that uses the three-dimensional shape data to measure the particle size and number of the granular material and calculates the particle size distribution of the granular material by assuming that the area ratio of the granular material is the weight ratio of the granular material.
[2] The particle size calculation unit calculates the particle size distribution of the granular material using the particle size, the number, and the following equations (1) and (2).

In the above equations (1) and (2), n _i is the number (pieces) of granules of particle size D _i , n _si is the measured value (pieces) of said number, α is a constant, D _i is the particle size (mm) of said granules, and W _i is the weight frequency of granules of particle size D _i .
[3] The particle size calculation unit of the granular material particle size measuring device described in [1] or [2] calculates the weight frequency of the granular material using the three-dimensional shape data when the three-dimensional shape data satisfies predetermined conditions, and does not calculate the weight frequency of the granular material using the three-dimensional shape data when the three-dimensional shape data does not satisfy predetermined conditions.
[4] The particle size calculation unit determines that the condition is satisfied if a normalized average height, calculated by dividing the average value of the heights of the three-dimensional shape data by the average particle size of the granular materials, is within a range of 3 to 8, and a normalized standard deviation, calculated by dividing the standard deviation of the heights by the average particle size of the granular materials, is within a range of 0.5 to 1.8, and determines that the condition is not satisfied if at least one of the normalized average height and the normalized standard deviation is outside the above range.
[5] A particle size measuring device for granular material described in [3] or [4], wherein the particle size calculation unit determines that the condition is satisfied when the average width direction of the standard deviation of the height of the three-dimensional shape data in the conveying direction is greater than a predetermined threshold value, and determines that the condition is not satisfied when the average width direction of the standard deviation of the height in the conveying direction is equal to or less than the threshold value.
[6] A particle size measuring device for granular material described in any of [3] to [5], wherein the particle size calculation unit determines that the condition is satisfied if the amount of change in the conveyor speed while creating the line profile data is within a predetermined range, and determines that the condition is not satisfied if the amount of change in the conveyor speed is outside the predetermined range.
[7] The particle size calculation unit inputs three-dimensional shape data created by the three-dimensional shape creation unit into a trained machine learning model, the input of which is three-dimensional shape data and the output of which is data indicating whether the condition is satisfied or not, and determines that the condition is satisfied if data indicating that the condition is satisfied is output from the trained machine learning model, and determines that the condition is not satisfied if data indicating that the condition is not satisfied is output from the trained machine learning model. This is a particle size measuring device for granular material described in [3].
[8] The particle size measuring device for granular material described in [7], wherein the machine learning model is a deep learning network that is machine-learned using multiple data sets, each set of which is a set of three-dimensional shape data and data indicating whether the three-dimensional shape data satisfies the condition, as training data.
[9] A method for measuring the particle size of a granular material, which measures the particle size distribution of a granular material loaded and transported on a conveyor, comprising: a line profile data creation step of creating line profile data of the granular material at a plurality of different positions in the transport direction using a line sensor; a three-dimensional shape data creation step of creating three-dimensional shape data of the granular material using the plurality of line profile data; and a particle size calculation step of measuring the particle size and number of the granular material using the three-dimensional shape data, and calculating the particle size distribution of the granular material by assuming that the area ratio of the granular material is the weight ratio of the granular material.
[10] The method for measuring the particle size of a granular material described in [9], wherein in the particle size calculation step, the particle size distribution of the granular material is calculated using the particle size, the number, and the following equations (1) and (2).

In the above equations (1) and (2), n _i is the number (pieces) of granules of particle size D _i , n _si is the measured value (pieces) of said number, α is a constant, D _i is the particle size (mm) of said granules, and W _i is the weight frequency of granules of particle size D _i .
[11] A method for measuring the particle size of a granular material described in [9] or [10], wherein in the particle size calculation step, if the three-dimensional shape data satisfies a predetermined condition, the three-dimensional shape data is used to calculate the weight frequency of the granular material, and if the three-dimensional shape data does not satisfy the predetermined condition, the weight frequency of the granular material is not calculated using the three-dimensional shape data.
[12] The method for measuring the particle size of a granular material described in [11], wherein in the particle size calculation step, the condition is satisfied if a normalized average height obtained by dividing the average value of the heights of the three-dimensional shape data by the average particle size of the granular material is within a range of 3 to 8 and a normalized standard deviation obtained by dividing the standard deviation of the heights by the average particle size of the granular material is within a range of 0.5 to 1.8, and the condition is not satisfied if at least one of the normalized average height and the normalized standard deviation is outside the above range.
[13] A method for measuring the particle size of a granular material described in [11] or [12], wherein in the particle size calculation step, the condition is satisfied when the average value of the width direction of the standard deviation of the height of the three-dimensional shape data in the conveying direction is greater than a predetermined threshold value, and the condition is not satisfied when the average value of the width direction of the standard deviation of the height in the conveying direction is equal to or less than the threshold value.
[14] A method for measuring the particle size of a granular material described in any one of [11] to [13], wherein in the particle size calculation step, the condition is satisfied if the amount of change in the speed of the conveyor while creating the line profile data is within a predetermined range, and the condition is not satisfied if the amount of change in the speed of the conveyor is outside the predetermined range.

The particle size measuring device and method of the present invention measure the particle size distribution online by taking into account not only the surface layer of the granular material loaded on a conveyor, but also the granular raw material present inside. This makes it possible to measure the particle size distribution of the granular material with higher accuracy than conventional particle size measuring devices that measure particle size using image data of the surface layer of the granular material loaded on a conveyor.

FIG. 1 is a schematic diagram showing the process from the production of briquettes to charging them into a coke oven. FIG. 2 is a schematic diagram showing a particle size measuring device 22 for a granular material according to this embodiment. FIG. 3 is a schematic diagram showing an example of the configuration of the image processing device 26. As shown in FIG. FIG. 4 is a schematic diagram for explaining a situation in which missing data occurs. FIG. 5 is an explanatory diagram showing an example of a method for complementing missing data. FIG. 6 shows three-dimensional shape images of the molded coal 16 before and after the interpolation process. FIG. 7 is a graph illustrating the process of identifying the boundaries of briquettes using the local minimum search method. FIG. 8 is an image showing the change in the three-dimensional shape data due to the process of identifying the boundaries of the briquettes. FIG. 9 is a diagram for explaining the probability of existence of piled-up granular objects in a region. FIG. 10 is a flow diagram showing an example of a method for measuring the particle size of a particulate object according to this embodiment. FIG. 11 is a schematic diagram for explaining a drop test of the molded coal 16 in the first embodiment.

The following describes a particle size measuring device and a particle size measuring method for granular material according to an embodiment of the present invention with reference to the drawings. Note that in the following embodiment, an example will be described in which the granular material is molded coal, but the granular material is not limited to molded coal, and the present invention can be applied to any granular material that is loaded and transported on a conveyor.

Figure 1 is a schematic diagram showing the process from when molded coal is produced to when it is charged into a coke oven. Raw coal 10, which is low-grade coal, is mixed with multiple materials including a binder by a mixer 12. The mixed raw materials mixed in the mixer 12 are molded into molded coal 16 in the form of briquettes by a molding machine 14. The molded coal 16 is loaded onto a conveyor 18 and transported toward a coke oven 20. The molded coal 16 is loaded into the coke oven 20 through a coal-charging hole and carbonized to produce coke. Note that while only one conveyor 18 is shown between the molding machine 14 and the coke oven 20 in Figure 1, in reality, the molded coal 16 is transported to the coke oven 20 via multiple conveyors 18.

Figure 2 is a schematic diagram showing a granular material particle size measuring device 22 according to this embodiment. As shown in Figure 2, the granular material particle size measuring device 22 is provided near the conveyor 18. The granular material particle size measuring device 22 includes a line sensor 24 and an image processing device 26. The line sensor 24 is, for example, a laser distance meter.

The line sensor 24 is provided above the conveyor 18. The line sensor 24 continuously irradiates the molded charcoal 16, which is transported at a constant speed by the conveyor 18, with a line-shaped laser light that runs along the width direction of the conveyor 18, and forms an image of the reflected light on an imaging element such as a CMOS. The line sensor 24 uses the image data formed on the imaging element to create line profile data that includes height information of the molded charcoal 16 at multiple positions in the width direction of the conveyor 18 that are different in the transport direction. The line sensor 24 outputs the multiple line profile data that have been created to the image processing device 26.

The image processing device 26 creates three-dimensional shape data of the molded coal 16 using the multiple line profile data created by the line sensor 24. If the three-dimensional shape data includes missing data, it is preferable that the image processing device 26 complements the missing data.

The image processing device 26 measures the particle size and number of the molded coal 16 using the three-dimensional shape data of the molded coal 16. The image processing device 26 calculates the particle size distribution of the molded coal 16 using the particle size and number of the molded coal 16, assuming that the area ratio of the granules is the weight ratio of the granules. Specifically, the image processing device 26 calculates the number of molded coal 16 using the particle size of the molded coal 16 and the following (1), and calculates the particle size distribution of the molded coal 16 using the particle size and number of the molded coal 16 and the following formula (2), assuming that the area ratio of the granules is the weight ratio of the granules.

上記（１）、（２）式において、ｎ_ｉは粒径Ｄ_ｉの成型炭１６の数（個）であり、ｎ_ｓｉは数の測定値（個）であり、αは定数であり、Ｄ_ｉは成型炭１６の粒径（ｍｍ）であり、Ｗ_ｉは粒径Ｄ_ｉの成型炭１６の重量度数である。ここで、成型炭１６の粒径とは、ヘイウッド径又は成型炭１６の大きさを矩形形状で近似した場合の縦横平均値である。

In the above formulas (1) and (2), n _i is the number (pieces) of coal briquettes 16 with a diameter D _i , n _si is the measured number (pieces), α is a constant, D _i is the diameter (mm) of the coal briquettes 16, and W _i is the weight fraction of the coal briquettes 16 with a diameter D _i . Here, the diameter of the coal briquettes 16 is the Heywood diameter or the average length and width when the size of the coal briquettes 16 is approximated by a rectangular shape.

When measuring the number of molded coals 16, it is preferable that the image processing device 26 identify the boundaries of the molded coals 16 in the three-dimensional shape data, binarize the three-dimensional shape data, and remove noise from the three-dimensional shape data. In this way, the particle size measuring device 22 of this embodiment measures the particle size of the molded coals 16 loaded and transported on the conveyor 18. Furthermore, it is preferable that the image processing device 26 measures the number of molded coals 16 after determining whether the three-dimensional shape data is suitable for measuring the number of molded coals 16.

Figure 3 is a schematic diagram showing an example of the configuration of the image processing device 26. The image processing device 26 is, for example, a general-purpose computer such as a workstation or a personal computer. The image processing device 26 has a control unit 28, an input unit 30, an output unit 32, and a memory unit 34. The control unit 28 is, for example, a CPU, and executes a program stored in the memory unit 34, causing the control unit 28 to function as a three-dimensional shape creation unit 36 and a grain size calculation unit 38.

The input unit 30 is, for example, a keyboard, a touch panel that is integrated with a display, etc. The output unit 32 is, for example, an LCD or CRT display, etc. The storage unit 34 is, for example, an updatable flash memory, a built-in hard disk or a hard disk connected via a data communication terminal, an information recording medium such as a memory card, and a read/write device for the information recording medium. The storage unit 34 stores programs for the control unit 28 to execute each function, data used by the programs, etc.

Next, the processing executed by the three-dimensional shape creation unit 36 and the particle size calculation unit 38 will be described. The three-dimensional shape creation unit 36 acquires a plurality of line profile data from the line sensor 24. The three-dimensional shape creation unit 36 converts the height information contained in the plurality of line profile data into 256 gradations of brightness values in the grayscale, and creates three-dimensional shape data of the molded coal 16 by arranging the plurality of line profile data converted into brightness values in the conveying direction. It is preferable that the three-dimensional shape creation unit 36 creates the three-dimensional shape data using a number of line profile data such that the length in the conveying direction is 0.5 m or more and 4.0 m or less.

The line profile data acquired from the line sensor 24 may contain missing data. FIG. 4 is a schematic diagram illustrating a situation in which missing data occurs. The line sensor 24 creates line profile data by irradiating the molded charcoal 16 with a line-shaped laser beam and forming an image of the reflected light reflected by the molded charcoal 16 on an imaging element. As shown in FIG. 4(a), depending on the surface condition of the molded charcoal 16 (e.g., the degree of surface gloss), there may be areas where the reflected light from the molded charcoal 16 does not form an image, and these areas become missing data. Also, as shown in FIG. 4(b), in the line sensor 24, the position where the laser beam is irradiated and the position where the reflected light is received are different, so that part of the reflected light is blocked by the molded charcoal 16, and there may be areas where the reflected light does not form an image, and these areas become missing data.

If the line profile data contains missing data, the three-dimensional shape data created from these data will also contain missing data. If the three-dimensional shape data contains missing data, the measurement accuracy will decrease when the three-dimensional shape data is used to measure the size and number of molded coals 16. For this reason, it is preferable that the three-dimensional shape creation unit 36 complements the missing data contained in the created three-dimensional shape data.

Figure 5 is an explanatory diagram showing an example of missing data complementation processing. For example, the three-dimensional shape creation unit 36 raster scans the three-dimensional shape data from left to right and top to bottom, and when missing data is detected, it complements it with the average luminance value of the surrounding eight pixels. In the example shown in Figure 5, the three-dimensional shape creation unit 36 complements pixel 0, which is missing data, with the average luminance value of pixels 1 to 4, which are not missing data, among the eight pixels (pixels 1 to 8 in Figure 5) surrounding pixel 0. Note that pixel 5 to the right of pixel 0 is also missing data, but the luminance value of pixel 0 complemented with the average luminance value of pixels 1 to 4 is also used to complement the missing data of pixel 5. The three-dimensional shape creation unit 36 performs such missing data complementation processing on the three-dimensional image data.

Figure 6 shows three-dimensional shape images of molded charcoal 16 before and after the interpolation process. Figure 6(a) shows a three-dimensional shape image of molded charcoal 16 before the missing data interpolation process. Figure 6(b) shows a three-dimensional shape image of molded charcoal 16 after the missing data interpolation process. By the above-mentioned interpolation process, the missing data scattered in the three-dimensional shape image of Figure 6(a) can be interpolated in a natural way, and three-dimensional shape data that does not include missing data as shown in Figure 6(b) can be obtained.

In the example shown in FIG. 5, the missing data is complemented with the average luminance value of the non-missing pixels among the eight pixels surrounding the missing data, but this is not limited to the above. The three-dimensional shape creation unit 36 may complement the missing data with the median luminance value, minimum luminance value, or maximum luminance value of the eight pixels surrounding the missing data, or may complement the missing data with the average luminance value or median luminance value excluding the maximum luminance value and minimum luminance value of the eight pixels surrounding the missing data. In addition, in order to eliminate abnormal values, a luminance value range may be set in advance based on past performance, and the missing data may be complemented using the average value or median value of the pixels within the set range. Furthermore, the order of raster scanning is not limited to left to right and top to bottom, but may be any of four combinations of right to left and up and down.

The three-dimensional shape creation unit 36 outputs the created three-dimensional shape data of the molded coal 16 to the particle size calculation unit 38. The particle size calculation unit 38 uses the three-dimensional shape data to measure the particle size and number of the molded coal 16, and uses these to calculate the particle size distribution of the molded coal 16. However, before that, it is preferable to identify the boundaries of the molded coal 16, binarize the three-dimensional shape data, and remove noise from the three-dimensional shape data. This makes it possible to improve the measurement accuracy when measuring the number of molded coals 16 using the three-dimensional shape data. These processes are described below.

First, the boundary identification process for identifying the boundary of the molded charcoal 16 will be described. FIG. 7 is a graph for explaining the process for identifying the boundary of the molded charcoal by the minimum value search method. FIG. 7(a) is a graph showing the valley detection method, and FIG. 7(b) is a graph showing the step detection method. As shown in FIG. 7(a), when the luminance value of the target pixel in the three-dimensional shape data satisfies "right tilt ≧ 0, left tilt ≦ 0" and "(right tilt - left tilt) ≧ threshold", the granularity calculation unit 38 detects that the target pixel is a valley. In addition, when the difference in luminance value between the target pixel and the neighboring pixel is greater than the threshold, the granularity calculation unit 38 detects that the target pixel is a step. The granularity calculation unit 38 identifies the target pixel of the detected valley and step as a boundary pixel of the molded charcoal 16, and changes the luminance value of the pixel to "0". Hereinafter, the process of changing the luminance value of the boundary pixel of the molded charcoal 16 to "0" will be referred to as "zero filling".

Figure 8 is an image showing the change in three-dimensional shape data due to processing to identify the boundaries of molded coal. The grain size calculation unit 38 vertically scans the three-dimensional shape data corresponding to the pre-processing image shown in Figure 8(a) to detect the boundary pixels of molded coal 16, and performs zero-filling of the pixels. The three-dimensional shape image in which the vertical boundary pixels have been zero-filled is shown in Figure 8(b). Similarly, the grain size calculation unit 38 horizontally scans the three-dimensional shape data corresponding to the pre-processing image shown in Figure 8(a) to detect the boundary pixels of molded coal 16, and performs zero-filling of the pixels. The three-dimensional shape image in which the horizontal boundary pixels have been zero-filled is shown in Figure 8(c).

The particle size calculation unit 38 combines the images of FIG. 8(b) and (c) by "OR operation" to create three-dimensional shape data in which the boundary pixels in the vertical and horizontal directions are filled with zeros. The three-dimensional shape image corresponding to this three-dimensional shape data is shown in FIG. 8(d). In this way, by performing a boundary identification process for identifying the boundary of the molded coal 16, the boundary of the molded coal 16 becomes clear, and the measurement accuracy of the number of molded coals 16 performed thereafter is improved. In the particle size measurement device 22 of the granular material according to this embodiment, the particle size calculation unit 38 identifies the boundary of the molded coal 16 using the minimum value search method, but this is not limited to this. The particle size calculation unit 38 may identify the boundary of the molded coal 16 using at least one of the minimum value search method, the Canny edge extraction method, the edge extraction method using a Sobel filter, and the edge extraction method using a Laplacian filter.

For example, when detecting the boundary of molded coal 16 using the Canny edge extraction method, the three-dimensional shape creation unit 36 performs differentiation processing on the three-dimensional shape data smoothed by a Gaussian filter using a Sobel filter or the like, and then determines whether or not it is the boundary of molded coal 16 by threshold processing using hysteresis. The three-dimensional shape creation unit 36 may also process the three-dimensional shape data using a Sobel filter or Laplacian filter, and then identify the boundary of molded coal 16 by threshold processing. In this way, by identifying the boundary of molded coal 16 and filling the boundary pixels with zero, the accuracy of the measurement of the number of molded coals 16 performed thereafter is improved.

Next, the binarization process will be described. The particle size calculation unit 38 preferably performs binarization processing on the three-dimensional shape data corresponding to the zero-filled image shown in FIG. 8(d). The threshold value used for binarization is determined in advance and stored in the memory unit 34. In this way, by performing binarization processing on the three-dimensional shape data, the accuracy of the measurement of the number of molded coals 16 performed thereafter is improved.

Next, the noise removal process will be described. It is preferable that the particle size calculation unit 38 removes noise from the three-dimensional shape data after the binarization process by performing at least one of the expansion process, contraction process, median filter, and averaging process at least once. By removing noise in this manner, the accuracy of the measurement of the number of molded coals 16 that is performed thereafter is improved.

上記（１）式において、ｎ_ｉは粒径Ｄ_ｉの成型炭１６の数（個）であり、ｎ_ｓｉは数の測定値（個）であり、αは定数であり、Ｄ_ｉは成型炭１６の粒径（ｍｍ）である。 The particle size calculation unit 38 performs area division by watershed on the three-dimensional shape data after the binarization process in which noise has been removed. This makes it possible to eliminate overlap between the briquettes 16, and to measure the number of each briquettes 16. The particle size calculation unit 38 measures the number of briquettes 16 included in the three-dimensional shape data using the area-divided three-dimensional shape data. Thereafter, the particle size calculation unit 38 calculates the number of briquettes 16 with a particle size D _{i using the particle size and number of the measured briquettes 16 and the following formula (1). Note that the particle size D i is} the particle size of any of the briquettes 16 included in the three-dimensional shape data, and is the particle size measured by the particle size calculation unit 38.

In the above formula (1), n _i is the number (pieces) of molded coal 16 having a particle size D _i , n _si is the measured number (pieces), α is a constant, and D _i is the particle size (mm) of the molded coal 16 .

上記（２）式において、ｎ_ｉは粒径Ｄ_ｉの成型炭１６の数（個）であり、ｎ_ｓｉは数の測定値（個）であり、Ｄ_ｉは成型炭１６の粒径（ｍｍ）であり、Ｗ_ｉは粒径Ｄ_ｉの成型炭１６の重量度数である After calculating the number of the coal briquettes 16 having a particle size D _i , the particle size calculation unit 38 calculates the weight percentage of the coal briquettes 16 having a particle size D _i by using the number and the following formula (2) assuming that the area ratio of the granules is the weight ratio of the granules. Here, the weight percentage is the weight ratio of the coal briquettes 16 having a particle size D _i .

In the above formula (2), n _i is the number (pieces) of coal briquettes 16 having a particle diameter D _i , n _si is the measured number (pieces), D _i is the particle diameter (mm) of the coal briquettes 16, and W _i is the weight frequency of the coal briquettes 16 having a particle diameter D _i.

The particle size calculation unit 38 calculates the weight frequency of all molded coal 16 included in the three-dimensional shape data. This allows the particle size calculation unit 38 to calculate the particle size distribution of the molded coal 16. The particle size calculation unit 38 may collectively calculate the particle sizes of the measured molded coal 16 in a predetermined range. For example, molded coal 16 with a particle size in the range of more than 3.0 mm and not more than 6.0 mm may be collectively measured as molded coal 16 with a particle size of 4.5 mm, and molded coal 16 with a particle size in the range of more than 6.0 mm and not more than 10.0 mm may be collectively measured as molded coal 16 with a particle size of 8.0 mm.

Here, we explain the above formulas (1) and (2). Figure 9 is a diagram that explains the probability of existence within an area of piled granular matter. Figure 9(a) is a top view of the piled granular matter, and Figure 9(b) is a side view.

9, based on the idea that the probability of piled granules existing in a given region is inversely proportional to the particle size, the number _ni of briquettes 16 with particle size _Di existing in a region of the three-dimensional shape data can be calculated by multiplying the measured number of briquettes 16 by (the reciprocal of particle size _Di )/(the sum of the reciprocals of all particle sizes). Based on this idea, the above formula (1) is used to calculate the number of briquettes 16 with an arbitrary particle size _Di.

Furthermore, since there is nothing other than the briquettes 16 on the conveyor 18, the weight frequency of the diameter D _i can be calculated by the volume ratio of the briquettes 16. The volume of the briquettes 16 with diameter D _i is calculated by (4/3) x π x (D _i /2) ³ x n _i , from which the above formula (2) is obtained. In this manner, the particle size calculation unit 38 calculates the weight frequency and particle size distribution of the briquettes 16. Since these weight frequencies and particle size distributions are only a portion of the briquettes 16, the particle size distribution of the briquettes 16 transported on the conveyor 18 can be obtained by using a large number of weight frequencies and particle size distributions obtained by repeatedly performing this process.

When the particle size distribution of the molded coal 16 is calculated, the particle size calculation unit 38 may display the calculated particle size and particle size distribution of the molded coal 16 on the output unit 32. This allows the operator to visually check the particle size and particle size distribution of the molded coal 16 displayed on the output unit 32 and thereby understand the particle size and particle size distribution of the molded coal 16.

In this way, the particle size calculation unit 38 calculates the weight frequency and particle size distribution of the molded coal 16. However, if the three-dimensional shape data acquired from the three-dimensional shape creation unit 36 is not suitable for calculating the particle size of the molded coal 16, it is preferable that the particle size calculation unit 38 does not use the weight frequency and particle size distribution of the molded coal 16 calculated using the three-dimensional shape data. For example, if the conveyor 18 includes a portion where the molded coal 16 is not loaded, it is preferable that the three-dimensional shape data is determined to be not suitable for calculating the weight frequency, and the particle size calculation unit 38 does not use the weight frequency and particle size distribution of the molded coal 16 calculated using the three-dimensional shape data. Conditions for determining whether the three-dimensional shape data is suitable for calculating the weight frequency are determined in advance, and the particle size calculation unit 38 determines whether the conditions are satisfied, thereby determining whether the weight frequency and particle size distribution of the molded coal 16 calculated using the three-dimensional shape data are used. Below, the conditions for determining whether the three-dimensional shape data is suitable for calculating the weight frequency are explained.

Whether or not the three-dimensional shape data includes a portion where the briquettes 16 are not loaded can be determined from the normalized average height obtained by dividing the average value of the heights of the three-dimensional shape data by the average particle diameter of the briquettes 16, and the normalized standard deviation obtained by dividing the standard deviation of the heights by the average particle diameter of the briquettes 16. Therefore, the particle size calculation unit 38 calculates the normalized average height using the average value of the heights of the three-dimensional shape data acquired from the three-dimensional shape creation unit 36 and the average particle diameter of the briquettes 16. In addition, the particle size calculation unit 38 calculates the normalized standard deviation using the heights of the three-dimensional shape data and the average particle diameter of the briquettes 16. The average particle diameter of the briquettes 16 can be calculated from the particle diameters D _i and the number of particles obtained using the three-dimensional shape data before the data is sorted. The average particle diameter of the briquettes 16 may be obtained using the particle diameters D _i and the number of particles measured for a predetermined period after the start of measurement.

If the normalized average height is within the range of 3 to 8 and the normalized standard deviation is within the range of 0.5 to 1.8, the particle size calculation unit 38 determines that the three-dimensional shape data is suitable for calculating the weight frequency, and uses the weight frequency and particle size distribution of the molded coal 16 calculated using the three-dimensional shape data. On the other hand, if the normalized average height is outside the range of 3 to 8 or the normalized standard deviation is outside the range of 0.5 to 1.8, the particle size calculation unit 38 determines that the three-dimensional shape data is not suitable for calculating the weight frequency, and does not use the weight frequency and particle size distribution of the molded coal 16 calculated using the three-dimensional shape data. This makes it possible to improve the calculation accuracy of the particle size of the molded coal 16 calculated by the particle size calculation unit 38. In the above example, a normalized average height obtained by dividing the average value of the height of the three-dimensional shape data by the average particle size of the molded coal 16 and a normalized standard deviation obtained by dividing the standard deviation of the three-dimensional shape data by the average particle size of the molded coal 16 were used, but this is not limiting and the judgment may be made using the average value of the height of the three-dimensional shape data and the standard deviation of the three-dimensional shape data.

Furthermore, when the conveyor 18 is stopped or when no molded coal 16 is loaded on the conveyor 18, the weight frequency of each particle size cannot be accurately calculated using the above formulas (1) and (2). Whether the conveyor 18 is stopped or whether no molded coal 16 is loaded on the conveyor 18 can be determined by whether the average value of the width direction of the standard deviation of the height of the three-dimensional shape data in the conveying direction is greater than a predetermined threshold value. Therefore, the particle size calculation unit 38 calculates the average value of the width direction of the standard deviation of the height of the three-dimensional shape data acquired from the three-dimensional shape creation unit 36 in the conveying direction. Then, when the value is greater than a predetermined threshold value, the particle size calculation unit 38 determines that the three-dimensional shape data is suitable for calculating the weight frequency, and uses the weight frequency and particle size distribution of the molded coal 16 calculated using the three-dimensional shape data. On the other hand, when the average widthwise value of the standard deviation of the height of the three-dimensional shape data in the conveying direction is equal to or less than a predetermined threshold value, the particle size calculation unit 38 determines that the three-dimensional shape data is not suitable for calculating the weight frequency, and does not use the weight frequency or particle size distribution of the molded coal 16 calculated using the three-dimensional shape data. This further improves the calculation accuracy of the weight frequency and particle size distribution of the molded coal 16 calculated by the particle size calculation unit 38.

The granularity calculation unit 38 may also determine whether the three-dimensional shape data is suitable for calculating weight frequency by inputting the three-dimensional shape data created by the three-dimensional shape creation unit 36 into a trained machine learning model in which the input is three-dimensional shape data and the output is data indicating whether the three-dimensional shape data is suitable for calculating weight frequency. For such trained machine learning model, it is preferable to use a deep learning network that has been machine-trained using as training data a large number of data sets in which the three-dimensional shape data and data indicating whether the three-dimensional shape data is suitable for calculating weight frequency. This makes it easy to determine whether the three-dimensional shape data is suitable for calculating weight frequency. This trained machine learning model is created in advance and stored in the storage unit 34.

It is also preferable that the molded coal 16 is transported at a constant speed by the conveyor 18. If the molded coal 16 is transported at a constant speed, the line sensor 24 creates line profile data at regular intervals in the transport direction, and the line profile data is used to create three-dimensional shape data that more accurately reflects the loading state of the molded coal 16. On the other hand, if the speed of the conveyor 18 changes while the line profile data is being created, the line profile data will not be created at regular intervals, and the three-dimensional shape data created using the line profile data will also be inaccurate.

Therefore, the particle size calculation unit 38 directly or indirectly acquires speed information from the conveyor 18 and determines whether the speed of the conveyor 18 has changed while the line profile data is being created. When the amount of change in the speed of the conveyor 18 while the line profile data is being created is within a predetermined range, the particle size calculation unit 38 determines that the speed of the conveyor 18 has not changed and that the three-dimensional shape data is suitable for calculating the weight frequency, and uses the weight frequency and particle size distribution of the molded coal 16 calculated using the three-dimensional shape data. On the other hand, when the amount of change in the speed of the conveyor 18 while the line profile data is being created is outside the predetermined range, the particle size calculation unit 38 determines that the speed of the conveyor 18 has changed and that the three-dimensional shape data is not suitable for calculating the weight frequency, and does not use the weight frequency and particle size distribution of the molded coal 16 calculated using the three-dimensional shape data. This further improves the calculation accuracy of the weight frequency and particle size distribution of the molded coal 16 calculated by the particle size calculation unit 38.

Figure 10 is a flow diagram showing an example of a method for measuring the particle size of granular material according to this embodiment. Next, the method for measuring the particle size of granular material according to this embodiment will be described with reference to Figure 10. The flow shown in Figure 10 is started, for example, when the particle size measuring device 22 for granular material is started and an input is received from the input unit 30 of the image processing device 26 instructing the start of measurement of the particle size distribution of the molded coal 16.

The line sensor 24 executes a line profile data creation step to irradiate the molded coal 16 with a line-shaped laser beam along the width direction of the conveyor 18, and creates line profile data including height information of the molded coal 16 at multiple different positions in the conveying direction (step S101). The line sensor 24 outputs the created line profile data to the three-dimensional shape creation unit 36 of the image processing device 26.

The three-dimensional shape creation unit 36 executes a three-dimensional shape data creation step to create three-dimensional shape data of the molded coal 16 using line profile data from multiple different positions in the conveying direction (step S102). It is preferable that the three-dimensional shape creation unit 36 complements missing data contained in the created three-dimensional shape data. The three-dimensional shape creation unit 36 outputs the created three-dimensional shape data to the particle size calculation unit 38.

The particle size calculation unit 38 measures the particle size and number of molded coals 16 using the three-dimensional shape data (step S103). In measuring the particle size and number of molded coals 16, it is preferable that the particle size calculation unit 38 executes a boundary identification process for identifying the boundaries of the molded coals 16, a binarization process, and a noise removal process.

上記（１）、（２）式において、ｎ_ｉは粒径Ｄ_ｉの成型炭１６の数（個）であり、ｎ_ｓｉは数の測定値（個）であり、αは定数であり、Ｄ_ｉは成型炭１６の粒径（ｍｍ）であり、Ｗ_ｉは粒径Ｄ_ｉの成型炭１６の重量度数である。粒度算出部３８は、取得した３次元形状データが重量度数の算出に適するか否かを判断し、重量度数の算出に適さないと判断した場合には、当該３次元形状データを用いて算出した重量度数や粒度分布を用いないことが好ましい。 The particle size calculation unit 38 measures the particle size and number of the briquettes 16 using the three-dimensional shape data, and then calculates the weight frequency and particle size distribution of the briquettes having a particle size _Di using the measured number and the following formulas (1) and (2) (step S104). The processes in steps S103 and S104 correspond to the particle size calculation step in the particle size measurement method for granular material.

In the above formulas (1) and (2), n _i is the number (pieces) of molded coal 16 with particle size D _i , n _si is the measured value of the number (pieces), α is a constant, D _i is the particle size (mm) of the molded coal 16, and W _i is the weight frequency of the molded coal 16 with particle size D _i . The particle size calculation unit 38 judges whether the acquired three-dimensional shape data is suitable for calculating the weight frequency, and if it judges that the data is not suitable for calculating the weight frequency, it is preferable not to use the weight frequency or particle size distribution calculated using the three-dimensional shape data.

The particle size calculation unit 38 displays the measured particle size distribution of the molded coal 16 on the output unit 32 (step S15). This allows the operator to visually confirm the particle size distribution of the molded coal 16 loaded on the conveyor and transported. Thereafter, the particle size calculation unit 38 judges whether an input instructing to end the calculation process of the particle size distribution of the molded coal 16 has been received via the input unit 30 (step S106). If the input has been received, the particle size calculation unit 38 judges that there is an end instruction (step S106: Yes) and ends the flow shown in FIG. 10. On the other hand, if there is no input indicating the end of the strength estimation of the molded coal 16, it judges that there is no end instruction (step S106: No), returns the process to step S101, and repeats the process from step S101 again.

In this way, the granular material particle size measuring device 22 and granular material particle size measuring method according to this embodiment measure the particle size distribution of the molded coal 16 online, taking into account not only the surface layer of the molded coal 16 loaded on the conveyor 18, but also the molded coal 16 present inside. Therefore, the particle size distribution of the granular material can be measured with higher accuracy than conventional particle size measuring devices that measure particle size from image data of the surface layer of the granular material loaded on the conveyor 18.

Example 1
Next, a description will be given of Example 1 in which the number of molded coals 16 after the drop test was measured using the particle size measuring device 22 for granular materials shown in FIG. 2. FIG. 11 is a schematic diagram for explaining the drop test of the molded coals 16 in Example 1. As shown in FIGS. 11(a) and 11(b), one to three drop tests were performed from a position of 2 m height. Thereafter, the molded coals were sieved with a sieve (mesh size 15×15 mm), and the molded coals sieved on the sieve were placed on a moving stage simulating a conveyor, and the number of molded coals was measured using the particle size measuring device 22 for granular materials. The imaging resolution in the width direction of the line sensor used was 0.3 mm/pixel, the imaging resolution in the conveying direction was 0.4 mm/pixel, and the moving speed of the moving stage was 200 mm/sec. In addition, missing data complementation processing, boundary identification processing, binarization processing, and noise removal processing were performed on the created three-dimensional image data. The measurement results of the number of molded coals are shown in Table 1 below.

In Table 1 above, rows 1-1 to 1-5, 2-1 to 2-5, and 3-1 to 3-5 show the results of five measurements taken after rearranging the samples on the moving stage. The "Number of briquettes" column shows the number of briquettes measured by the granular material particle size measuring device 22. The "Actual number" column shows the number of briquettes measured visually. As shown in Table 1, the number of briquettes and the actual number are almost the same, confirming that the number of briquettes can be measured with high accuracy by the granular material particle size measuring device 22 according to this embodiment.

Example 2
Next, Example 2 will be described in which the weight frequency of molded coal was calculated using the particle size measuring device 22 for granular materials shown in Figure 2 on the conveyor of the molded coal production line. Using a line sensor with an imaging resolution of 0.3 mm/pixel in both the vertical and horizontal directions, line profile data of molded coal loaded in about three layers on a conveyor transporting the coal at a speed of 240 m/min was created. Note that the zero level of the height was set by creating profile data of a conveyor on which no molded coal was loaded before the measurement.

In Example 1, three-dimensional shape data was created using the created line profile data. The created three-dimensional shape data was subjected to missing data completion processing, boundary identification processing, binarization processing, and noise removal processing, and then the particle size and number of molded coal were measured. These measurement results and the above formulas (1) and (2) were used to calculate the weight frequency of each particle size. The calculated weight frequency of each particle size was then used to calculate the frequency distribution curve of the molded coal.

In Example 2, three-dimensional shape data was created using line profile data created under the same conditions as in Example 1. The created three-dimensional shape data was subjected to missing data completion processing, boundary identification processing, binarization processing, and noise removal processing. If the normalized average height obtained by dividing the average value of the height of this three-dimensional shape data by the average particle size of the molded coal was within the range of 3 to 8, and the normalized standard deviation obtained by dividing the standard deviation of the three-dimensional shape data by the average particle size of the molded coal was within the range of 0.5 to 1.8, the three-dimensional shape data was determined to be suitable for calculating weight frequency, and the weight frequency of each particle size calculated using the three-dimensional shape data was used. The average particle size of the molded coal was 10 mm. The frequency distribution curve of the molded coal was then calculated using the weight frequencies of each particle size.

On the other hand, if the normalized average height, calculated by dividing the average height of the three-dimensional shape data by the average particle size of the molded coal, is outside the range of 3 to 8, or if the normalized standard deviation, calculated by dividing the standard deviation of the three-dimensional shape data by the average particle size of the molded coal, is outside the range of 0.5 to 1.8, the three-dimensional shape data is determined to be unsuitable for calculating weight frequency, and the weight frequency of each particle size calculated using the three-dimensional shape data is discarded without being used.

In Example 3, three-dimensional shape data was created using line profile data created under the same conditions as Example 1. The created three-dimensional shape data was subjected to missing data complementation, boundary identification, binarization, and noise removal. If the normalized average height obtained by dividing the average value of the height of this three-dimensional shape data by the average particle size of the molded coal was within the range of 3 to 8, the normalized standard deviation obtained by dividing the standard deviation of the three-dimensional shape data by the average particle size of the molded coal was within the range of 0.5 to 1.8, and the average value of the standard deviation in the conveying direction in the width direction was greater than 10, it was determined that the three-dimensional shape data was suitable for calculating the weight frequency, and the weight frequency of each particle size calculated using the three-dimensional shape data was used. The weight frequency of each particle size was then used to calculate the frequency distribution curve of the molded coal.

On the other hand, if the normalized average height, calculated by dividing the average height of the three-dimensional shape data by the average particle size of the molded coal, is outside the range of 3 to 8, if the normalized standard deviation, calculated by dividing the standard deviation of the three-dimensional shape data by the average particle size of the molded coal, is outside the range of 0.5 to 1.8, or if the average value of the standard deviation in the conveying direction in the width direction is 10 or less, then the three-dimensional shape data is determined to be unsuitable for calculating the weight frequency, and the weight frequency of each particle size calculated using the three-dimensional shape data is discarded without being used.

In Example 4, three-dimensional shape data was created using line profile data created under the same conditions as in Example 1. The created three-dimensional shape data was subjected to missing data complementation, boundary identification, binarization, and noise removal. If the normalized average height obtained by dividing the average value of the height of this three-dimensional shape data by the average particle size of the molded coal was within the range of 3 to 8, the normalized standard deviation obtained by dividing the standard deviation of the three-dimensional shape data by the average particle size of the molded coal was within the range of 0.5 to 1.8, the average value of the standard deviation in the conveying direction in the width direction was greater than 10, and the change in the speed of the conveyor while the line profile data was created was within ±10%, the three-dimensional shape data was determined to be suitable for calculating the weight frequency, and the weight frequency of each particle size calculated using the three-dimensional shape data was used. Then, the weight frequency of each particle size was used to calculate the frequency distribution curve of the molded coal.

On the other hand, if the normalized average height obtained by dividing the average height of the three-dimensional shape data by the average particle size of the molded coal is outside the range of 3 to 8, the normalized standard deviation obtained by dividing the standard deviation of the three-dimensional shape data by the average particle size of the molded coal is outside the range of 0.5 to 1.8, the average widthwise value of the standard deviation in the conveying direction is 10 or less, or the change in conveyor speed during the creation of the line profile data exceeds ±10%, the three-dimensional shape data is determined to be unsuitable for calculating the weight frequency, and the weight frequency of each particle size calculated using the three-dimensional shape data is discarded without being used. Note that, although the above example shows an example in which the change in speed of the conveyor 18 is determined to be within ±10%, it is more preferable to determine the change in conveyor speed within ±3%.

In Comparative Example 1, instead of a line sensor, image data captured by an area camera with an imaging resolution of 0.25 mm/pixel in both the vertical and horizontal directions was analyzed to measure the particle size and number of molded coal surface layers, and the molded coal weight frequency was calculated using the molded coal volume ratio calculated from these measurement results. The molded coal frequency distribution curve was then calculated using the calculated weight frequency of each particle size.

The correlation between the particle size frequency distribution curves of Examples 1 to 4 and Comparative Example 1 and the particle size frequency distribution curves actually measured by sieving the same molded coal (hereinafter referred to as the "actually measured frequency distribution curves") was confirmed. The results are shown in Table 2 below.

As shown in Table 2, the coefficient of determination (R ² ) between the frequency distribution curve of the molded coal calculated from the weight frequency calculated in Examples 1 to 4 and the actually measured frequency distribution curve was 0.80 or more, and it was confirmed that the weight frequency of the molded coal could be calculated with high accuracy. In addition, it was confirmed that the accuracy of the particle size measurement of the molded coal was improved by determining whether or not the three-dimensional shape data is suitable for calculating the weight frequency, and not using the weight frequency data calculated from the three-dimensional shape data when it is determined that the three-dimensional shape data is not suitable for calculating the weight frequency.

On the other hand, the coefficient of determination ( ^R2 ) between the molded coal frequency distribution curve calculated from the weight frequency measured in Comparative Example 1 and the actual measured frequency distribution curve was 0.43, and the molded coal weight frequency could not be calculated with high accuracy. It is difficult to identify the boundaries of molded coal in the image data captured by the area camera, and this is thought to have reduced the measurement accuracy of the particle size and number of molded coal, and therefore the measurement accuracy of the weight frequency. Furthermore, in Comparative Example 1, the internal state of the molded coal loaded on the conveyor is not taken into consideration, and this is thought to have further reduced the measurement accuracy of the molded coal weight frequency.

Reference Signs List 10 Raw coal 12 Mixer 14 Molding machine 16 Molded coal 18 Conveyor 20 Coke oven 22 Granular material particle size measuring device 24 Line sensor 26 Image processing device 28 Control unit 30 Input unit 32 Output unit 34 Memory unit 36 Three-dimensional shape creation unit 38 Particle size calculation unit

Claims (22)

A particle size measuring device for measuring a particle size distribution of a granular material loaded and transported on a conveyor, comprising:
a line sensor that creates line profile data of the granular object at a plurality of different positions in a conveying direction;
An image processing device,
the image processing device includes a three-dimensional shape creation unit that creates three-dimensional shape data of the granular object using a plurality of line profile data created by the line sensor;
a particle size calculation unit that uses the three-dimensional shape data to measure the particle size and number of the granular objects and calculates the particle size distribution of the granular objects by assuming that the area ratio of the granular objects is the weight ratio of the granular objects. The particle size measuring device for a granular material according to claim 1 , wherein the particle size calculation unit calculates a particle size distribution of the granular material by using the particle size, the number, and the following equations (1) and (2).

In the above equations (1) and (2), n _i is the number (pieces) of granules of particle size D _i , n _si is the measured value (pieces) of said number, α is a constant, D _i is the particle size (mm) of said granules, and W _i is the weight frequency of granules of particle size D _i . The particle size measuring device for granular material according to claim 1 or claim 2, wherein the particle size calculation unit calculates the weight frequency of the granular material using the three-dimensional shape data when the three-dimensional shape data satisfies a predetermined condition, and does not calculate the weight frequency of the granular material using the three-dimensional shape data when the three-dimensional shape data does not satisfy the predetermined condition. The particle size measuring device of claim 3, wherein the particle size calculation unit determines that the condition is satisfied if a normalized average height obtained by dividing the average value of the heights of the three-dimensional shape data by the average particle size of the granular material is within a range of 3 to 8, and a normalized standard deviation obtained by dividing the standard deviation of the heights by the average particle size of the granular material is within a range of 0.5 to 1.8, and determines that the condition is not satisfied if at least one of the normalized average height and the normalized standard deviation is outside the above range. The particle size measuring device for granular material according to claim 3, wherein the particle size calculation unit determines that the condition is satisfied when the average width of the standard deviation of the height of the three-dimensional shape data in the conveying direction is greater than a predetermined threshold value, and determines that the condition is not satisfied when the average width of the standard deviation of the height in the conveying direction is equal to or less than the threshold value. The particle size measuring device for granular material according to claim 4, wherein the particle size calculation unit determines that the condition is satisfied when the average value of the width direction of the standard deviation of the height of the three-dimensional shape data in the conveying direction is greater than a predetermined threshold value, and determines that the condition is not satisfied when the average value of the width direction of the standard deviation of the height in the conveying direction is equal to or less than the threshold value. The particle size measuring device for granular material according to claim 3, wherein the particle size calculation unit determines that the condition is satisfied if the amount of change in the speed of the conveyor while creating the line profile data is within a predetermined range, and determines that the condition is not satisfied if the amount of change in the speed of the conveyor is outside the predetermined range. The particle size measuring device for granular material according to claim 4, wherein the particle size calculation unit determines that the condition is satisfied if the amount of change in the speed of the conveyor while creating the line profile data is within a predetermined range, and determines that the condition is not satisfied if the amount of change in the speed of the conveyor is outside the predetermined range. The particle size measuring device for granular material according to claim 5, wherein the particle size calculation unit determines that the condition is satisfied if the amount of change in the speed of the conveyor while creating the line profile data is within a predetermined range, and determines that the condition is not satisfied if the amount of change in the speed of the conveyor is outside the predetermined range. The particle size measuring device for granular material according to claim 6, wherein the particle size calculation unit determines that the condition is satisfied if the amount of change in the speed of the conveyor while creating the line profile data is within a predetermined range, and determines that the condition is not satisfied if the amount of change in the speed of the conveyor is outside the predetermined range. The granularity calculation unit inputs the three-dimensional shape data created by the three-dimensional shape creation unit into a trained machine learning model, the trained machine learning model being an input of the three-dimensional shape data and an output of data indicating whether or not the condition is satisfied,
A particle size measuring device for granular matter as described in claim 3, which determines that the condition is satisfied when data indicating that the condition is satisfied is output from the trained machine learning model, and determines that the condition is not satisfied when data indicating that the condition is not satisfied is output from the trained machine learning model. The particle size measuring device for granular material according to claim 11, wherein the machine learning model is a deep learning network that is machine-learned using multiple data sets, each set being a set of three-dimensional shape data and data indicating whether the three-dimensional shape data satisfies the condition, as training data. A method for measuring the particle size distribution of granular objects loaded and transported on a conveyor, comprising the steps of:
a line profile data creation step of creating line profile data of the granular object at a plurality of different positions in a conveying direction using a line sensor;
a three-dimensional shape data creation step of creating three-dimensional shape data of the granular object using a plurality of line profile data;
a particle size calculation step of measuring the particle size and number of the granular objects using the three-dimensional shape data, and calculating a particle size distribution of the granular objects by assuming that the area ratio of the granular objects is the weight ratio of the granular objects;
A method for measuring the particle size of a granular material, comprising the steps of: The method for measuring the particle size of a granular material according to claim 13, wherein in the particle size calculation step, a particle size distribution of the granular material is calculated using the particle size, the number, and the following formulas (1) and (2).

In the above equations (1) and (2), n _i is the number (pieces) of granules of particle size D _i , n _si is the measured value (pieces) of said number, α is a constant, D _i is the particle size (mm) of said granules, and W _i is the weight frequency of granules of particle size D _i . The method for measuring the particle size of a granular material according to claim 13 or 14, wherein in the particle size calculation step, if the three-dimensional shape data satisfies a predetermined condition, the three-dimensional shape data is used to calculate the weight frequency of the granular material, and if the three-dimensional shape data does not satisfy the predetermined condition, the three-dimensional shape data is not used to calculate the weight frequency of the granular material. The method for measuring the particle size of a granular material according to claim 15, wherein in the particle size calculation step, the condition is satisfied if a normalized average height obtained by dividing the average value of the heights of the three-dimensional shape data by the average particle size of the granular material is within a range of 3 to 8, and a normalized standard deviation obtained by dividing the standard deviation of the heights by the average particle size of the granular material is within a range of 0.5 to 1.8, and the condition is not satisfied if at least one of the normalized average height and the normalized standard deviation is outside the above range. The method for measuring the particle size of a granular material according to claim 15, wherein in the particle size calculation step, the condition is satisfied when the average value of the width direction of the standard deviation of the height of the three-dimensional shape data in the conveying direction is greater than a predetermined threshold value, and the condition is not satisfied when the average value of the width direction of the standard deviation of the height in the conveying direction is equal to or less than the threshold value. The method for measuring the particle size of a granular material according to claim 16, wherein in the particle size calculation step, the condition is satisfied when the average value of the width direction of the standard deviation of the height of the three-dimensional shape data in the conveying direction is greater than a predetermined threshold value, and the condition is not satisfied when the average value of the width direction of the standard deviation of the height in the conveying direction is equal to or less than the threshold value. The method for measuring the particle size of a granular material according to claim 15, wherein in the particle size calculation step, the condition is satisfied if the amount of change in the speed of the conveyor while creating the line profile data is within a predetermined range, and the condition is not satisfied if the amount of change in the speed of the conveyor is outside the predetermined range. The method for measuring the particle size of a granular material according to claim 16, wherein in the particle size calculation step, the condition is satisfied if the amount of change in the speed of the conveyor while creating the line profile data is within a predetermined range, and the condition is not satisfied if the amount of change in the speed of the conveyor is outside the predetermined range. The method for measuring the particle size of a granular material according to claim 17, wherein in the particle size calculation step, the condition is satisfied if the amount of change in the speed of the conveyor while creating the line profile data is within a predetermined range, and the condition is not satisfied if the amount of change in the speed of the conveyor is outside the predetermined range. The method for measuring the particle size of a granular material according to claim 18, wherein in the particle size calculation step, the condition is satisfied if the amount of change in the speed of the conveyor while creating the line profile data is within a predetermined range, and the condition is not satisfied if the amount of change in the speed of the conveyor is outside the predetermined range.