Device and method for bulk material analysis and bulk material production system

Abstract

The invention relates to a method (100) and a device (20) for bulk material analysis and to a bulk material production system (10) and a computer program product. According to the invention, individual bulk material bodies (16) of a bulk material (12) are identified (S2) in a 3D data set (14). For each identified bulk material body (16), a compensation calculation is carried out (S3), in which the bulk material body (16) is reconstructed from the 3D data set (14) on the basis of data (D) associated with the bulk material body (16). A bulk material characteristic (C) is then determined (S4) on the basis of in each case at least one parameter value which is obtained in the compensation calculations and characterizes the relevant identified bulk material body (16).

Description

[0001] Description

[0002] Device and method for bulk material analysis and bulk material production system

[0003] field of technology

[0004] The invention relates to a method and a device for bulk material analysis as well as a bulk material production system and a computer program product.

[0005] State of the art

[0006] When producing bulk materials such as pellets from a large number of individual bulk material particles, also referred to as "grains," a high degree of homogeneity is typically desired. This particularly applies to the shape and / or size of the bulk material particles.

[0007] For example, metal ore pellets used in direct reduction furnaces can have stringent requirements regarding size distribution. Pellets that are too large or too small are unsuitable for the direct reduction process. Pellets that are too large can lead to poor metallurgical reduction, while pellets that are too small can negatively impact the gas flow through the furnace. Therefore, a pellet size distribution that is as ideally suited as possible to the subsequent (reduction) process should be achieved during pellet production.

[0008] It is therefore often essential to determine the size distribution of bulk solids in the bulk material, which can provide information about the degree of homogeneity or the deviation from a desired degree of homogeneity.

[0009] One way to determine such a size distribution is sieve analysis. This involves sieving the bulk material through sieves with different mesh sizes. The different size fractions can then be quantified accordingly. Sieve analysis is therefore a mechanical analysis. While this type of sieve analysis is robust, it is complex and time-consuming.

[0010] As an alternative to such mechanical analyses, size distributions can also be determined using extractive methods for laboratory investigations or in-line analysis methods. For example, it is known to evaluate images of the bulk material using digital, i.e., computer-aided, image processing. In this case, appropriate algorithms and, as described in Deo et al., "Machine Learning based Image Processing for Iron

[0011] Ore Pellet Size Analysis," 2021 International Conference on Nascent Technologies in Engineering, possibly supported by machine learning, the individual bulk solids in the image are detected. The area occupied by the detected bulk solids in the image can then provide an indication of the size of these bulk solids.

[0012] However, such image-based analysis methods also pose problems. For example, partially obscured bulk solids can distort the size distribution. Furthermore, it is often difficult to reliably identify individual bulk solids in the image, especially in poor lighting conditions and the associated low contrast. This is especially true if such bulk solids analysis is to be carried out not in a laboratory but in an industrial environment, such as a production facility.

[0013] From Andersson et al., “Pellet Size Estimation Using Spherical Fitting”, 2007 IEEE Instrumentation and Measurement Technology Conference, and Thurley et al., “An industrial 3D Vision system for size measurement of iron ore green pellets using morphological image segmentation”, Minerals Engineering 21 (2008) 405-415, it is also known to record 3D surface data of pellets and to identify individual pellets by segmentation.

[0014] Against this background, it is an object of the present invention to improve bulk material analysis, in particular to enable a robust and at the same time low-effort characterization of bulk material.

[0015] This object is achieved by the method and the device for bulk material analysis as well as the bulk material production system and the computer program product according to the independent claims.

[0016] Preferred embodiments are the subject of the dependent claims and the following description.

[0017] Summary of the invention

[0018] According to a first aspect of the invention, in the computer-implemented method for bulk material analysis, in particular for the analysis of metal ore pellets for direct reduction furnaces, individual bulk material bodies of a bulk material are identified in a 3D data set. For each identified bulk material body, a reconstruction calculation is performed, in which the bulk material body is reconstructed, in particular virtually, in particular three-dimensionally, based on data assigned to the bulk material body from the 3D data set. Based on at least one parameter value obtained in the reconstruction calculations, which characterizes the respective identified bulk material body, a bulk material characteristic for the bulk material is then determined. A 3D data set within the meaning of the present invention is preferably a data set containing 3D information. A 3D data set can also be referred to as a spatial data set.The 3D data set can, for example, characterize a spatial arrangement of the bulk material bodies and / or a spatial characteristic of the bulk material bodies, i.e., their plasticity or three-dimensional shape. The 3D data set expediently contains discrete values ​​of Cartesian coordinates, i.e., XYZ coordinates, that characterize at least parts of bulk material bodies, such as surface sections. The 3D data set preferably represents a three-dimensional point cloud, with each data point representing a surface of an object, for example, a bulk material body. A 3D data set can, for example, be formed from a plurality of elevation profiles.

[0019] A bulk material characteristic within the meaning of the invention is preferably a description of the bulk material by one or more variables. A bulk material characteristic can, for example, contain a distribution of parameter values ​​of a parameter that characterize individual bulk material bodies, such as a size distribution of the bulk material bodies. However, it is also conceivable that a bulk material characteristic is given by a single parameter value, such as an average size of the bulk material bodies.

[0020] One aspect of the invention is based on the approach of evaluating 3D information on a bulk material, for example metal ore pellets for direct reduction furnaces, in order to determine the properties of the bulk material. For this purpose, a 3D data set is expediently provided, such as can result from a three-dimensional recording of the bulk material. In contrast to conventional methods, in which only two-dimensional images of the bulk material - effectively projections of the bulk material in a plane - are analyzed, the proposed method can draw on additional information relating to the third dimension, for example height profiles. In addition to identifying individual bulk material bodies in the 3D data set, a compensation calculation can be carried out in which the individual bulk material bodies are virtually reconstructed, i.e., virtually reproduced.Since data from the 3D data set assigned to the respective bulk solid bodies is used for the reconstruction, a three-dimensional reconstruction can be carried out. This virtual reconstruction expediently provides at least one parameter value that can be used to characterize the corresponding (real) bulk solid body. A large number of such parameter values ​​then allows the determination of a bulk solid characteristic. For example, the parameter values ​​determined as part of the adjustment calculation can be statistically evaluated. Compared to conventional analysis methods, the adjustment calculation provides sufficient (3D) data to enable a robust (three-dimensional) reconstruction or simulation of the bulk solid body, even for partially hidden bulk solid bodies.Therefore, even with poor separation, a larger number of bulk material bodies can be correctly "measured" per data set, thus determining representative bulk material characteristics. The bulk material characteristics determined in this way are preferably output for further use, for example, via an interface. For example, the bulk material characteristics can be output to a control program and thus serve as the basis for plant control. Alternatively or additionally, the bulk material characteristics can also be output to a user, for example, displayed on a screen.

[0021] Typically, the individual bulk solids of a bulk material can be assumed to be, at least for the most part, essentially spherical, i.e., at least approximately spherical. Within the scope of the fitting calculation, spheres can then be reconstructed from this data, for example, whose (virtual) radius or diameter can then be assigned to the corresponding bulk solids. The data from the 3D dataset assigned to the identified bulk solids can thus, for example, characterize a spherical cap of the respective bulk solid.

[0022] Preferred embodiments of the invention and their further developments are described below. These embodiments can be combined with each other and with the aspects of the invention described below, unless expressly excluded.

[0023] One possible approach for reconstructing the identified bulk solids based on the data assigned to them from the 3D dataset is to adapt a, in particular predetermined, geometric body to this data as part of the adjustment calculation. This geometric body expediently represents a, in particular three-dimensional, model of the bulk solid. As part of the adjustment calculation, the size of this geometric body can then, for example, be adapted to the data assigned to the respective bulk solid. The computing power required to adapt a, in particular predetermined, geometric body to data from the 3D dataset is usually manageable. Furthermore, a bulk solid can be reliably and robustly reconstructed using just a small amount of data.For example, just 50 data points corresponding to a surface of the bulk solid may be sufficient for a reliable and robust reconstruction.

[0024] The adaptation of a given geometric body to the data from the 3D data set assigned to the respective bulk material body can also be understood as "fitting" the geometric body to the data. In other words, the data assigned to the respective bulk material body are expediently fitted during the adjustment calculation, e.g. with a model of the bulk material body, in order to be able to determine at least one parameter value characterizing the bulk material body, such as its diameter or radius. Since, as already noted above, the individual bulk materials of a bulk material are usually at least approximately - i.e. within the scope of a first approximation - spherical or can be assumed to be spherical, it is expedient to fit a spherical body, i.e. a sphere, to the data assigned to the respective bulk material body during the adjustment calculation. As a result, the adjustment calculation can, for example,Four parameter values ​​are determined: the position of the center of the spherical body, such as an X, a Y, and a Z coordinate, and its radius or diameter. The radius or diameter thus determined, i.e., adjusted to the respective data, can then be assigned to the corresponding real bulk solid body as a characterizing parameter value.

[0025] It may be helpful and / or advantageous for the characterization of the bulk solid to determine possible deviations from the assumed spherical shape. A parameter value characterizing the bulk solid is determined on the basis of a large number of errors that are determined for the data assigned to the bulk solid from the 3D data set, in particular for each data point assigned to the bulk solid, as part of the adjustment calculation. A spatial distribution of the errors is taken into account. This makes it possible to determine, for example, how homogeneous the fit of the geometric body to the data is or how strongly the fit deviates locally from the data. This can be used to conclude that the bulk solid is out of round. This means that out of round can be determined as a parameter value characterizing the bulk solid.

[0026] The errors can, for example, represent a deviation of the reconstruction from the actual bulk solid body. The errors can therefore, for example, be a measure of the deviation of the individual data or data points from the adjusted, i.e., fitted, geometric body. If the bulk solid body under consideration essentially corresponds exactly to the geometric body, the determined errors will essentially represent noise in terms of their spatial distribution. This means that the errors are spatially uncorrelated or homogeneously distributed over the surface of the (virtual) body. This homogeneity of the spatial error distribution disappears, however, if the shape of the bulk solid body does not correspond to the geometric body, at least in some sections.

[0027] It is therefore conceivable, for example, to group the defects, i.e. to summarize them in groups. Each defect group can provide an indication of a local deviation of a section of the bulk material body from the geometric body. For this purpose, it is expedient to group the defects on the basis of a spatial correlation and to determine the parameter value characterizing the bulk material body on the basis of this defect grouping. The defect groups can be used to identify protrusions or indentations, for example. Alternatively or additionally, the data assigned to the respective bulk material body can be divided into several data sets. A parameter value characterizing the bulk material body can then be determined from the errors determined in connection with the several data sets as part of the adjustment calculation.For example, such a parameter value can be determined from several errors that arise when reconstructing the bulk solid body and represent deviations of the reconstruction from the data from the multiple data sets. For each data set, a measure of the agreement with the (virtual) reconstructed body can be determined. For example, it can be determined how well the data points from the various data sets lie on the surface of the (virtual) reconstructed body. Conveniently, an average error is determined for each data set.

[0028] Preferably, a ratio of the errors, especially the average ones, is determined, and the parameter value characterizing the bulk solid is derived from this. This not only allows the quality of the bulk solid reconstruction or the adjustment calculation to be verified, but also allows the determination of any out-of-roundness of the bulk solid.

[0029] The multiple data sets preferably each contain data representing a spherical zone, i.e., the surface of one spherical layer, of the bulk material. For spherical bulk materials, the resulting reconstruction error is the same for all data sets. However, if the shape deviates from sphericality, a higher error is to be expected for some of the data sets. Thus, simply dividing the data into two or three data sets may be sufficient to reliably determine the out-of-roundness.

[0030] An efficient method for reconstructing the respective bulk solid body from the data assigned to it in the fitting calculation is based on linear algebra. Accordingly, a matrix equation is set up and solved for the fitting calculation, preferably on the basis of the data assigned to the respective bulk solid body. It is then advantageous to determine at least one parameter value characterizing the respective bulk solid body, for example its radius or diameter, from a solution vector of the matrix equation. The matrix equation can represent a system of equations whose number of unknowns depends on the geometric body chosen as the model for the bulk solid body. If, for example, an attempt is made in the fitting calculation to fit a spherical body, i.e. a sphere, to the data assigned to the bulk solid body, orTo fit, the system of equations can contain four unknowns: three Cartesian coordinates for the center and a radius or diameter.

[0031] Depending on the amount of data assigned to the bulk solid, for example, the number of corresponding data points, the system of equations is overdetermined. It is therefore advisable to solve the matrix equation using the least squares method. This is not only advantageous in terms of the required computing power, but can also provide an error or quality factor for the reconstruction performed during the adjustment calculation for each identified bulk solid. This quality factor, as a characterizing parameter value, can also form the basis for determining the bulk solid's characteristics.

[0032] A 3D dataset suitable for bulk material analysis can be generated using a profile sensor, past which the bulk material is moved at a known speed. Consequently, such a 3D dataset can be generated anywhere where a continuous bulk material flow is (already) present and / or the movement speed of the bulk material can be determined or known. For example, such a dataset can be generated on a transport line between a device for generating bulk material bodies and a sorting device where bulk material bodies that are too small and / or too large are sorted out, and the bulk material analysis can be performed accordingly.

[0033] The profile sensor preferably has a camera that detects distortions in a line projected onto the bulk material by a projection device, e.g., a laser. A height profile can be derived from these distortions. Such a profile sensor can thus precisely detect the bulk material even under adverse environmental conditions, such as poor lighting. Several such height profiles, which are recorded successively as the bulk material continues to be transported past the profile sensor, i.e., through a detection range of the profile sensor, in particular, through the field of view of the camera, preferably form a 3D data set. This data set then depicts the topology of the bulk material moving past the profile sensor.

[0034] This approach can result in shadowing, which prevents the camera from capturing at least portions of the projected laser line. Consequently, such shadowing can lead to "blind spots" in 3D space. This effect can be reduced by projecting multiple laser lines and / or using multiple cameras.

[0035] In order to perform a s / t assessment of the bulk material production, i.e., an "online measurement," the profile, particularly the height profile, of the bulk material moving past the profile sensor is recorded essentially continuously. For this purpose, the bulk material can be recorded at a high frequency, for example, in the kilohertz range. In particular, the distortions of the line projected onto the bulk material can be recorded at this high frequency. With the resulting "gapless" recording of the bulk material, the bulk material characteristics can be determined essentially in real time. This allows, in particular, online monitoring of the quality of the produced bulk material.

[0036] This also makes it possible to generate multiple 3D data sets consecutively. These data sets are preferably recorded within a predetermined time interval and are therefore of the same size, which can increase comparability and facilitate evaluation. Preferably, several of these 3D data sets, each recorded within the predetermined time interval, are evaluated sequentially. This makes it possible to monitor the development of the bulk material characteristics and, consequently, the production process.

[0037] Particularly effective production of bulk material with specified properties, for example, a predetermined average bulk material body or grain size and / or a predetermined size distribution, can be achieved by utilizing this development of the bulk material characteristics. For example, individual production parameters can be adjusted if the bulk material characteristics change. Generally speaking, a bulk material production process is preferably controlled based on the determined bulk material characteristics.

[0038] If necessary, individual process steps can also be carried out using AI. Preferably, the identification of individual bulk material bodies is carried out using artificial intelligence. It has been shown that this allows for reliable identification of bulk material bodies even under changing environmental conditions. Consequently, AI-based identification of bulk material bodies is particularly advantageous for a continuously running system in which bulk material is to be conveyed along a conveyor line and analyzed in the process.

[0039] According to a second aspect of the invention, a device for bulk material analysis, in particular for analyzing metal ore pellets for direct reduction furnaces, comprises: i) an identification means for identifying individual bulk material bodies of a bulk material in a 3D data set; ii) a computing means for performing a compensation calculation for each identified bulk material body, in which the bulk material body is reconstructed, in particular virtually, in particular three-dimensionally, on the basis of data assigned to the bulk material body from the 3D data set; and iii) an analysis means for determining a bulk material characteristic for the bulk material on the basis of at least one parameter value obtained in each case during the compensation calculations, which parameter value characterizes the respective identified bulk material body.

[0040] Because the reconstruction of the identified bulk material bodies is based on data from a 3D data set, a particularly robust determination of at least one parameter value characterizing the respective bulk material body can be carried out. Such reconstruction is even possible in cases where the data from the 3D data set is incomplete, i.e. only depicts the respective bulk material body in sections. The device is therefore particularly suitable for bulk material analysis in cases where the bulk material is not or only slightly isolated. In particular, such a device can also be used to analyze bulk material during the transport of bulk material piled up on a conveyor belt, for example immediately after its production.

[0041] A means within the meaning of the present invention can be embodied in hardware and / or software. The means can in particular have a processing unit, preferably one that is data- or signal-connected to a memory and / or bus system. For example, the means can have a microprocessor unit (CPU) or a module thereof and / or one or more programs or program modules. The means can be designed to process instructions implemented as a program stored in a memory system, to detect input signals from a data bus and / or to output output signals to a data bus. A memory system can have one or more, in particular different, storage media, in particular optical, magnetic, solid-state and / or other non-volatile media. The program can be designed in such a way that it embodies the methods described here orcapable of carrying out such a process, so that the means can carry out the steps of such processes and thus, in particular, carry out a bulk material analysis.

[0042] According to a third aspect of the invention, the bulk material production system, in particular a system for producing metal ore pellets for direct reduction furnaces, comprises a device according to the second aspect of the invention and a device for bulk material production, in particular at least one pelletizing plate. The bulk material production system can also comprise a profile sensor for detecting the produced bulk material, in particular for detecting height profiles of the bulk material. The profile sensor is expediently arranged in the region of a conveying device for conveying the produced bulk material after its production, for example, from at least one pelletizing plate to a sorting device. This allows the quality of the produced bulk material to be determined and monitored in situ or online.

[0043] A fourth aspect of the invention relates to a computer program product for carrying out the method according to the first aspect of the invention. Such a computer program product can be executed, for example, on a control unit of a bulk material production system. The computer program product expediently contains instructions which, when executed by a computer, such as a control unit of the bulk material production system, cause the computer to carry out the method according to the first aspect of the invention.

[0044] Brief Description of the Drawings The above-described properties, features, and advantages of this invention, as well as the manner in which they are achieved, will become clearer and more readily understood in connection with the following description of an embodiment, which is explained in more detail in conjunction with the drawings.

[0045] Fig 1 an example of a bulk material production system;

[0046] Fig 2A shows an example of a 3D data set in a two-dimensional representation;

[0047] Fig 2B shows bulk solids identified in the 3D dataset from Fig 2A;

[0048] Fig 3 an example of a fitting calculation in which a bulk solid body is reconstructed;

[0049] Fig 4A shows an example of data from a 3D data set divided into multiple data sets and assigned to a bulk material body;

[0050] Fig 4B shows the data from Fig 4A in a side view;

[0051] Fig 5 an example of a bulk material characteristic;

[0052] Fig. 6 shows an example of a method for bulk material analysis; and

[0053] Fig. 7 shows an example of a grouping of errors obtained in a fitting calculation for the reconstruction of a bulk solid body based on a spatial correlation.

[0054] Where appropriate, the same reference numerals are used in the figures for the same or corresponding elements of the invention.

[0055] Description of the embodiments

[0056] FIG. 1 shows an example of a bulk material production system 10 with a device 20 for bulk material analysis and a device 30 for bulk material production, in particular pellet production. In the present example, the bulk material 12 produced from feedstock X, in particular pellets, can be ejected from a production unit 32 of the device 30, designed here as a pelletizing plate, onto a transport device 34. If necessary, further production units (not shown) can also be provided upstream of the production unit 32 with respect to a transport direction R of the transport device 34, which then also eject the produced bulk material 12 onto the transport device 34. By means of the transport device 34, the produced bulk material 12 can then be continuously transported away, for example, to a sorting device 36.In this case, the bulk material 12 expediently passes through a sensor system 28, which can be assigned to the device 20 for bulk material analysis. The properties of the bulk material 12 can thus be determined essentially continuously and, consequently, based on the entire bulk material 12 detectable by the sensor system 28. In particular, the bulk material production can be monitored online in this way.

[0057] In the present example, the device 20 for bulk material analysis comprises a profile sensor 22 forming the sensor system 28, with a projection unit 24 and a camera 26. The projection unit 24, for example a laser, is expediently configured to project a line 40 transversely to the transport direction R (and therefore only shown as a point in FIG. 1) onto the bulk material 12 moving past the profile sensor 22. The camera 26 is configured to detect the distortions of this line 40 resulting from the projection onto the (uneven) bulk material 12. In particular, the distortions of the line 40 resulting from the movement of the bulk material 12 past the profile sensor 22 can also be detected in this way. The device 20 can generate a 3D data set from the detected distortions, for example by means of a data processing unit (not shown). This 3D data set can depict the formation of the bulk material 12 at the grain level, iethe size, shape and / or relative position of the individual grains or bulk solids.

[0058] The device 20 for bulk material analysis further comprises an identification means 20a for identifying individual bulk material bodies of the bulk material 12 in the 3D data set and a computing means 20b for performing a compensation calculation for each identified bulk material body. The computing means 20b is configured to reconstruct each identified bulk material body based on data from the 3D data set that is associated with the respective bulk material body. Additionally, an analysis means 20c is provided for determining a bulk material characteristic based on at least one parameter value obtained during the compensation calculations, which characterizes the respective identified bulk material body.

[0059] FIG. 2A shows an example of a 3D data set 14 in a two-dimensional representation, in which individual bulk material bodies 16 can be identified. For reasons of clarity, only one of the bulk material bodies 16 is provided with a reference symbol. The 3D data set 14 is expediently composed of a plurality of height profiles, each of which was recorded along a horizontal line in FIG. 2A (cf. reference symbol 40 in FIG. 1).

[0060] FIG. 2B shows bulk material bodies 16 identified in the 3D data set 14 from FIG. 2A, which are identified by different hatching in the representation shown. Here, too, for reasons of clarity, only one of the identified bulk material bodies 16 is marked with a reference symbol. The identification of the individual bulk material bodies 16 from the 3D data set 14, which can also be referred to as "segmentation," is expediently carried out using the two-dimensional representation of the 3D data set 14 shown in FIG. 2A or a similar two-dimensional representation. Conventional digital image processing algorithms known per se can then be used for the identification.

[0061] Preferably, during identification, data from the 3D data set 14 is assigned to each bulk material body 16. With these data, at least a portion of each individual identified bulk material body 16 can be displayed and / or analyzed three-dimensionally.

[0062] FIG. 3 shows an example of a compensation calculation in which a bulk material body 16 is reconstructed. Data D from a 3D data set is assigned to the bulk material body. This data D can, for example, be in the form of a point cloud. The individual data points of the point cloud are then expediently located in the area of ​​a surface of the bulk material body 16, in particular within a section of the surface that can be captured, for example, by the camera 26 from FIG. 1 and is not shadowed by other bulk material bodies. The individual data points of the point cloud can therefore characterize at least part of the surface of the bulk material body 16.

[0063] For the reconstruction of the bulk material body 16, a predetermined geometric body 18, in the present example a spherical body, i.e. a sphere, is expediently adapted to the data D. For example, the radius or diameter of the body 18 and its position in space, e.g. the position of its center point, can be adapted to the data D until the data D lie in the area of ​​the surface of the body 18. In other words, the geometric body 18 can be fitted to the data D. The geometric body 18 consequently expediently forms a model for the real bulk material body 16. In this way, a (virtual) reconstruction, in particular a three-dimensional one, of the bulk material body 16 is obtainable.

[0064] Mathematically, the adjustment calculation involves setting up a matrix equation and solving it, for example, using the method of least (error) squares.

[0065] If, as in the present example, the adjustment calculation is based on a spherical body with the center (x0,y0,z0) in Cartesian coordinates and the radius r, the following applies:

[0066] (x - x0) 2 + (y - y0) 2 + (z ~ zo) 2 = r 2 .

[0067] After extension and rearrangement, this equation is x 2 + y 2 + z 2 = 2xx0+ 2yy0+ 2zz0+ r 2 — XQ — y(j — z(j . For data points from the data D, the matrix equation f = Ac can be set up, where

[0068] The Cartesian coordinates x t , y t and z twith i = 1, 2, ..., n, and thus the vector / and the matrix A are completely known. The system of equations can therefore be solved for n > 4. When identifying bulk solid bodies 16 in a 3D data set, a lower limit is preferably set for the number of data points, e.g., n > 50. This actually overdetermines the system of equations / = Ac. The solution vector c, and thus the radius and center of the spherical body 18, can then be calculated using the least squares method.

[0069] FIG 4A shows an example of data D from a 3D data set, divided into multiple data sets D1, D2, and D3, assigned to a bulk solid. As shown in FIG 3, the data D contains a point cloud of multiple data points located within the area of ​​a surface of the identified bulk solid. In this example, the data D represents a spherical cap of the bulk solid. FIG 4A shows the distribution of the data points when viewed perpendicular to this spherical cap. From this perspective, the data points lie within a circle.

[0070] In this representation, data sets D1, D2, and D3 contain data points from annuli of this circle, indicated by the dashed lines. In three dimensions, these annuli each correspond to the surface of a spherical layer, i.e., a so-called spherical zone. Consequently, data sets D1, D2, and D3 each represent a spherical zone of the bulk solid.

[0071] If, for example, a spherical body is fitted to the data D, as shown in FIG 3, the non-circularity of the bulk material body can be derived from the determined mean errors for the deviation of the data points of each data set D1, D2, D3 from the body fitted to the data D. This is clearly shown in FIG 4B:

[0072] FIG 4B shows the data D from FIG 4A as a solid line and a spherical body 18 fitted to the data D in a side view. In this view, the data D form a circular segment marked by hatching, which is divided into layers according to the data sets D1, D2, D3. The deviations of the data D from the body 18 increase towards the apex of the circular segment. The ratios of the mean deviations of data points from the individual data sets D1, D2, D3 to the body 18, i.e. the ratios of the mean errors determined for the individual data sets D1, D2, D3 when fitting the body 18 to the data D, characterize the out-of-roundness of the actual bulk material body.

[0073] FIG 5 shows an example of a bulk material characteristic C. The bulk material characteristic C is, purely by way of example, a distribution 42 of the bulk material body or grain size, i.e., the diameter of the bulk material bodies of a bulk material. This distribution 42 can be determined by creating a histogram from all determined diameters of bulk material bodies identified in a 3D data set.

[0074] In principle, a direct conversion of a size distribution into a volume distribution is also possible. By converting and summing the individual bulk solid volumes from different size ranges, statistical bulk solid values ​​can be determined, similar to the results of sieve analyses.

[0075] In addition to the distribution 42, the bulk material characteristic C can also contain an average bulk material size 44, which is shown in FIG 5 as a dashed line.

[0076] FIG 6 shows an example of a method 100 for bulk material analysis.

[0077] In process step S1, a 3D data set associated with a bulk material is provided. For this purpose, the bulk material can be detected using sensors in process step S1, for example, with a profile sensor.

[0078] In a method step S2, individual bulk solids of the bulk material are identified in the 3D data set. As explained above in connection with FIGS. 2A, 2B, this is expediently done in a two-dimensional representation of the 3D data set. This two-dimensional representation can also be generated in method step S2, in addition to the actual identification of the individual bulk solids. For this purpose, for example, data points from the 3D data set can be projected onto a plane and coded with regard to height information, i.e., their distance from the plane. Based on this coding, the data points can be segmented and each assigned to a bulk solid.

[0079] In a method step S3, a fitting calculation is performed for each identified bulk solid. During this fitting calculation, the bulk solid is reconstructed based on the data assigned to it from the 3D data set. For example, a geometric body, such as a sphere, can be adjusted or fitted to the data assigned to the bulk solid, as described above in connection with FIG. 3. During this fitting calculation, at least one parameter value characterizing the bulk solid can be determined, for example, its radius or diameter.

[0080] Alternatively or additionally, in process step S3, the out-of-roundness of the bulk material body can also be determined as a characterizing parameter value. For this purpose, the data are conveniently divided into several data sets, as explained above in connection with Figures 4A and 4B, and the mean errors resulting from the adjustment calculation for each of these data sets are compared.

[0081] In a method step S4, a bulk material characteristic is derived from at least one parameter value obtained in the adjustment calculations, which characterizes the respective identified bulk material body. For example, a distribution of the parameter values ​​obtained in the adjustment calculations can be determined. Alternatively or additionally, the obtained parameter values ​​can be averaged to an average value or used as the basis for another statistical method. In principle, it is conceivable that the bulk material characteristic contains several such distributions or (average) values.

[0082] The determined bulk material characteristics are expediently also output in method step S4, for example, via a corresponding interface. The bulk material characteristics can be output, in particular, to a system controller configured to regulate a bulk material production system. The system controller can control the bulk material production system, in particular at least one device for bulk material production, in a further method step S5 based on the determined bulk material characteristics, for example, by changing production parameters. With the aid of such a control system, the quality of the produced bulk material can be increased or reliably maintained at a predetermined quality level.

[0083] FIG 7 shows an example of groups G1, G2 of errors F obtained in a fitting calculation for the reconstruction of a bulk solid. The grouping is based on a spatial correlation of the errors F. For clarity, only some of the errors F are provided with a reference symbol.

[0084] In the present example, the errors F represent the deviations from data D assigned to a bulk material body or data points from a 3D data set to a predetermined geometric body 18. The magnitude of the deviation is represented here by the hatching. Unhatched points represent small errors F, i.e., only slight positive or negative deviations, while diagonally hatched points represent larger positive deviations and cross-hatched points represent larger negative deviations. Those points that lie outside the geometric body 18 exhibit a positive deviation, while those points that lie within the geometric body 18 exhibit a negative deviation.

[0085] According to the purely exemplary representation in FIG. 7, there is a cluster of larger positive deviations in one region of the geometric body 18 and a cluster of larger negative deviations in another region. The corresponding errors F are grouped into groups G1 and G2 according to this spatial correlation.

[0086] Corresponding to this error grouping, the data D can also be grouped. Based on the error grouping, it is therefore possible not only to identify the proportion of "deviating" data (points), but also to identify in detail which data (points) deviate significantly from the predetermined geometric body 18.

[0087] Although the invention has been illustrated and described in detail by the preferred embodiments, the invention is not limited by the disclosed examples and other variations may be derived therefrom by those skilled in the art without departing from the scope of the invention.

[0088] List of reference symbols

[0089] 10 Bulk material production system

[0090] 12 Bulk goods

[0091] 14 3D dataset

[0092] 16 bulk solids

[0093] 18 geometric bodies

[0094] 20 Device for bulk material analysis

[0095] 20a Identification means

[0096] 20b Calculating tools

[0097] 20c Analysis medium

[0098] 22 Profile sensor

[0099] 24 Projection unit

[0100] 26 Camera

[0101] 28 Sensor technology

[0102] 30 Device for bulk material production

[0103] 32 production units

[0104] 34 Transport device

[0105] 36 Sorting device

[0106] 40 Line

[0107] 42 Distribution

[0108] 44 average bulk material size

[0109] 100 methods for bulk solids analysis

[0110] S1 Provision of 3D data set

[0111] S2 Identification of bulk solids

[0112] S3 balancing calculation

[0113] S4 Determining a bulk material characteristic

[0114] S5 Control of bulk material production

[0115] C Bulk material characteristics

[0116] D Data

[0117] D1, D2, D3 dataset

[0118] X Input material

[0119] R Transport direction

[0120] F Error G1, G2 Group

Claims

Claims 1. Computer-implemented method (100) for bulk material analysis, wherein - individual bulk material bodies (16) of a bulk material (12) are identified in a 3D data set (14) (S2), - for each identified bulk material body (16), a compensation calculation is carried out (S3), in which the bulk material body (16) is reconstructed on the basis of data (D) assigned to the bulk material body (16) from the 3D data set (14), and - on the basis of at least one parameter value obtained in the adjustment calculations, which characterizes the respective identified bulk material body (16), a bulk material characteristic (C) for the bulk material (12) is determined (S4), wherein a parameter value characterizing the bulk material body (16) is determined on the basis of a spatial distribution of a plurality of errors which are determined for the data (D) assigned to the bulk material body (16) from the 3D data set (14) in the context of the adjustment calculation.

2. Method (100) according to claim 1, wherein, within the scope of the adjustment calculation (S3), a, in particular predetermined, geometric body (18) is adapted to data (D) from the 3D data set (14) which are assigned to the bulk material body (16).

3. Method (100) according to claim 2, wherein, within the scope of the adjustment calculation (S3), a spherical body is adapted to the data (D) associated with the respective bulk material body (16).

4. Method (100) according to one of the preceding claims, wherein the defects are grouped on the basis of a spatial correlation and the parameter value characterizing the bulk material body (16) is determined on the basis of this defect grouping.

5. Method (100) according to one of the preceding claims, wherein the data (D) assigned to the respective bulk material body (16) are divided into several data sets (D1, D2, D3) and for each data set (D1, D2, D3) a measure of the correspondence with the reconstructed bulk material body is determined.

6. Method (100) according to one of the preceding claims, wherein for the compensation calculation (S3) a matrix equation is set up and solved on the basis of the data (D) assigned to the respective bulk material body (16), so that at least one parameter value characterizing the respective bulk material body (16) can be derived from a solution vector of the matrix equation.

7. Method (100) according to one of the preceding claims, wherein the 3D data set (14) is generated by means of a profile sensor (22) past which the bulk material (12) is moved at a known speed.

8. The method (100) according to claim 7, wherein a profile, in particular a height profile, of the bulk material (12) moving past the profile sensor (22) is recorded substantially continuously and the bulk material characteristic (C) is determined substantially in real time.

9. Method (100) according to one of the preceding claims, wherein a plurality of 3D data sets (14) each recorded within a predetermined time interval are evaluated sequentially.

10. Method (100) according to one of the preceding claims, wherein a bulk material production process is controlled (S5) on the basis of the determined bulk material characteristic (C).

11. Method (100) according to one of the preceding claims, wherein the identification of the individual bulk material bodies (16) is carried out using artificial intelligence.

12. The method (100) according to any one of the preceding claims, wherein the bulk material bodies are metal ore pellets for direct reduction furnaces and the bulk material characteristics are determined during pellet production.

13. Device (20) for bulk material analysis, comprising - an identification means (20a) for identifying individual bulk material bodies (16) of a bulk material (12) in a 3D data set (14), - a computing means (20b) for carrying out a compensation calculation for each identified bulk material body (16), in which the bulk material body (16) is reconstructed on the basis of data (D) assigned to the bulk material body (16) from the 3D data set (14), and - an analysis means (20c) for determining a bulk material characteristic (C) for the bulk material (12) on the basis of at least one parameter value obtained in the adjustment calculations, which characterizes the respective identified bulk material body (16), wherein a parameter value characterizing the bulk material body (16) is determined on the basis of a spatial distribution of a plurality of errors which are determined for the data (D) assigned to the bulk material body (16) from the 3D data set (14) in the context of the adjustment calculation.

14. Bulk material production system (10), comprising a device (20) for bulk material analysis according to claim 13 and a device (30) for bulk material production.

15. Computer program product for executing a method (100) according to one of claims 1 to 12.

Images