Method for selecting a dosing device and determining operating parameters for dosing a bulk material with unknown flow properties

Abstract

In a method for selecting a dosing device (1) and determining operating parameters for dosing a bulk material (2) with unknown flow properties using this dosing device (1), an optical image of a predetermined quantity of the bulk material (2) is acquired in a reference environment (4). Based on the optical image, geometric parameters of the bulk material (2) are determined using a geometry evaluation module of an electronic data processing device. A user is guided through a dialogue with an input module of an electronic data processing device, answering a number of questions by means of a binary decision.Using a machine learning method and based on the determined geometric parameters and the user's responses, a characteristic value for the flow properties of the bulk material (2) is determined with a characteristic value evaluation module of an electronic data processing device. The user enters a desired conveying rate for the bulk material (2) via the input module and receives a suggestion for a suitable dosing device (1) as well as for suitable operating parameters for dosing the bulk material (2), which are based on the determined characteristic value for the flow properties and the desired conveying rate.

Description

Technical field

[0001] The present invention relates to a method for selecting a dosing device and for determining suitable operating parameters for dosing a bulk material with initially unknown flow properties. In particular, the invention relates to a method based on a combination of image-based analysis, user interaction, and machine learning to automatically determine the optimal dosing device and the corresponding operating parameters. State of the art

[0002] Various methods for selecting dosing devices for bulk materials are known in industry.

[0003] Traditionally, dosing devices, such as screw conveyors, vibratory feeders, or belt conveyors, are selected based on known physical properties of the bulk material, such as particle size, moisture content, density, and flow behavior. These physical properties typically require time-consuming, manual laboratory testing before a suitable dosing device can be selected and the corresponding operating parameters defined.

[0004] However, a known problem with these methods is that users often find it difficult to determine the flow properties of a bulk material, especially if it is a novel or previously untested material. In many cases, users have to resort to empirical testing, manually trying out and adjusting various dosing devices until a suitable solution is found. This approach is time-consuming and resource-intensive, as it requires repeated adjustments of operating parameters and multiple test series.

[0005] Furthermore, the flow properties of bulk solids can be influenced by external factors such as humidity, temperature, and particle size distribution, making it even more difficult for the user to select the right dosing device and the optimal operating parameters. Precise, user-friendly procedures that simplify the process of selecting and configuring a dosing device for a bulk solid with initially unknown properties are often lacking.

[0006] Known approaches to automating this process are generally based on extensive datasets containing pre-determined flow properties of various bulk materials. However, these databases only offer added value if the physical properties of the bulk material are already known and there is an exact match with an existing sample from the database. In practice, however, it is often found that such a match is not always present, especially if the bulk material exhibits deviations in composition or physical properties.

[0007] For the user, this means that they often have to rely on numerous experimental tests to find the appropriate dosing device and its operating parameters. This manual approach is not only time-consuming but also carries the risk of suboptimal settings, which can lead to inefficient processes or faulty dosing results.

[0008] It is also known from practice that the user obtains various pieces of information about a bulk material with previously unknown flow properties and, using an evaluation module of an electronic data processing device, determines an estimated characteristic value for the flow properties based on the previously entered information. Based on this value, a suitable dosing device and suitable operating parameters for dosing the bulk material with the respective dosing device are then determined and specified. For example, EP3859287A1 also discloses the use of sensors or additional material measuring devices to determine the physical properties of the raw material in order to increase dosing accuracy.The information to be determined by the user in advance can, for example, include a photographic image of a given quantity of the bulk material in a reference environment, on the basis of which geometric parameters of the bulk material can then be determined, which are also taken into account for determining the characteristic value for the flow properties.

[0009] For example, DE 10 2019 104 293 A1 discloses a method for configuring an agricultural spreading machine in which an image of the granular material to be spread is evaluated in such a way that a suitable configuration of a rotary-driven conveying device is selected based on the determined material properties. Similarly, EP2924417A2 discloses a method for calculating the particle size of fertilizer using a photographic image, which can be used to determine a recommended setting for a fertilizer spreader.

[0010] However, it has been shown that determining the characteristic value for the flow properties solely on the basis of a photograph leads in most cases to unsatisfactory results when specifying the dosing device and the operating parameters.

[0011] Therefore, further information is required, which the user must obtain and enter. The more detailed the information and the more specific or precise this information must be defined and entered by the user beforehand, the better the results; however, the effort required to obtain the information also increases, and the more error-prone the process becomes if individual pieces of information are obtained or specified incorrectly by the user. Summary of the invention

[0012] It is therefore considered an object of the present invention to quickly and reliably determine the suitable dosing device and the optimal operating parameters for a bulk material with initially unknown flow properties.

[0013] This problem is solved according to the invention by a method for selecting a metering device and determining operating parameters for metering a bulk material with unknown flow properties, comprising the steps of: creating an optical image of a predetermined quantity of the bulk material in a reference environment; determining geometric parameters of the bulk material based on the optical image using a geometry evaluation module of an electronic data processing device; guiding a user through a dialogue using an input module of an electronic data processing device, wherein the user answers a number of questions by means of a binary decision; determining a characteristic value for the flow properties of the bulk material using a characteristic value evaluation module of an electronic data processing device with the aid of a machine learning method, based on the determined geometric parameters and the user's answers; Users enter a desired delivery rate via the input module; Proposing a suitable dosing device and appropriate operating parameters for this dosing device for dosing the bulk material, based on the determined characteristic value for the flow properties and the desired conveying rate.

[0014] An optical image can be created by a user with minimal effort. It is advantageous to use a predefined reference environment whose relevant properties are known to the geometry evaluation module or provided to it. In the simplest case, the reference environment can be a flat reference surface with a uniform color that differs from the surroundings and whose size is known. For the most precise evaluation of the optical image, it can be advantageous for the reference environment to provide at least a one-dimensional scale, or a two-dimensional scale in a plane, and preferably a three-dimensional scale, which can be used by the geometry evaluation module to determine geometric parameters for individual particles of the bulk material as well as for an accumulation of the bulk material as reliably and precisely as possible.The reference environment can, for example, be a planar reference surface on which reference markings, preferably with a scale division, are applied or displayed in two orthogonal coordinate directions. In addition to the planar reference surface, the reference environment can also have a side wall projecting from the planar reference surface along at least one edge, on which reference markings, preferably with a scale division, are also applied or displayed. Any reasonably planar and horizontally oriented surface can also be used as the reference environment, with the reference markings being created by the user, for example, by adding a ruler or scale, which are visible in the optical representation.

[0015] The additional information the user must enter consists solely of a number of binary decisions in response to predefined questions, with the binary decisions being queried via a dialog. The dialog and the questions posed can be designed in such a way that the user can answer them very reliably without the use of additional aids such as measuring devices or experimental setups. The dialog can be completed by the user within a short time, for example, within a few minutes or even seconds.

[0016] The method according to the invention enables the efficient and accurate selection of not only a suitable dosing device for the bulk material in question, but also the determination of the operating parameters to be set for the desired dosing of the bulk material with this dosing device, largely automatically and without the otherwise necessary complex investigations. By providing a dialog in which only binary decisions from the user are expected and recorded, the information required for determining the dosing device and its operating parameters can be easily and intuitively acquired.Using a suitable predefined machine learning method, the characteristic value for the flow properties of the bulk material and, based on this, the proposal for the dosing device and its operating parameters can be determined with increasing precision, starting from just a few and easy-to-answer questions in the course of the dialogue, without increasing the effort required for the individual user.

[0017] An optical image encompasses any type of visual representation generated by optical means. This could include photographic images as well as other types of optical measurements or representations, such as images from infrared cameras, scanner images, or even holographic representations. Preferably, the optical image is a photograph of the bulk material taken by a camera on a mobile device. Most users already have access to a mobile device with a sufficiently high-quality camera. With a suitably designed software program, which can be easily accessed and executed on a mobile device, the geometry evaluation module, the input module, and the parameter evaluation module, as well as the machine learning process, can also be implemented on the mobile device.Apart from the mobile device, no further aids are required for the user, so the procedure can be carried out almost anytime, anywhere and without much effort.

[0018] The geometric parameters can include all parameters derivable from the optical image without additional user input, in particular the size and shape of individual particles of the bulk material, as well as the height of the accumulated bulk material, the volume of the accumulation, and / or the angles of the cone surface. These parameters are directly linked to the flow properties of the bulk material, enabling precise determination of the appropriate dosing device.

[0019] The metering device is advantageously selected from a group of known metering devices, including screw conveyors, vibratory feeders, or belt conveyors. For this group of known metering devices, the individual operating parameters and their permissible ranges can be predefined and stored for retrieval. Selecting from a known group of metering devices accelerates the decision-making process and reduces complexity, allowing for good results even with just a few questions. The method is flexible and can be applied to various metering devices, enabling broad industrial application.

[0020] For some applications, several dosing devices can be used, each capable of conveying and dosing the respective bulk material at the desired conveying rate. These dosing devices regularly differ in several parameters and characteristics, as well as in their respective purchase costs and the operating costs typically expected.To enable and achieve the user-desired dosing accuracy and, at the same time, to suggest the most cost-effective dosing device from among various available options, an advantageous embodiment of the invention provides that, in addition to the desired feed rate, the user enters a desired dosing accuracy via the input module. Based on the determined flow properties, the desired feed rate, and the desired dosing accuracy, a suitable dosing device and appropriate operating parameters for this device are then suggested for dosing the bulk material. By inputting and considering the desired dosing accuracy, individual user requirements can be better accommodated.Depending on the dosing device, the dosing accuracy can be specified either as a single value or a desired range of values, or as a combination of several independent parameters or characteristics that define the dosing accuracy, such as values ​​or ranges for dosing consistency and a maximum tolerable dosing error. In addition to or alongside the dosing accuracy, other parameters or properties of the available dosing devices can also be entered by the user and considered for the automated selection and specification of suitable operating parameters.

[0021] Optionally, the operating parameters of the dosing device can include conveying speed, dosing frequency, and / or tilt angle. These parameters can be adjusted to the specific properties of the bulk material to achieve optimal results. Precise definition of these parameters leads to efficient and uniform dosing of the bulk material. Additional operating parameters, potentially specific to individual dosing devices, can also be defined and determined using this method.

[0022] According to an advantageous embodiment of the invention, the questions in the dialogue relate to physical properties of the bulk material, such as particle size, moisture content, or surface roughness. During the dialogue, the user can also be asked to provide information on other properties, such as strength, shape, deformability, or three-dimensional surface structuring of individual particles of the bulk material. The user is guided by targeted questions, which simplifies the data acquisition process. By focusing on relevant physical properties, the accuracy of the analysis is increased.

[0023] According to a particularly advantageous embodiment of the invention, the dialogue can be conducted via a graphical user interface. The questions posed in the graphical user interface present the user with two images illustrating contrasting properties of the bulk material. The user then selects, by means of a binary decision, which image best describes the property of the bulk material they have provided, as exemplified by the two images. The individual questions can be formulated using descriptive terms and illustrated with intuitively understandable images, thus facilitating the user's decision-making process.A significant advantage of the method according to the invention is that the user only needs to make binary decisions to answer the questions during the dialogue, and then, based on these binary decisions, not only can a suitable dosing device be selected and suggested, but operating parameters can also be determined and specified, which do not necessarily have to be represented by binary values, but can assume any values ​​within a permissible range for the dosing device in question. Thus, the method according to the invention also performs a transformation of a data format from binary decisions and inputs to real values ​​within a predetermined range.

[0024] To ensure that the dialog process is as simple and intuitive as possible while still eliciting meaningful information from the user, it can be implemented that individual questions are repeated multiple times, each time accompanied by two different images representing contrasting properties. For example, the question "How solid is the bulk material?" can be repeated several times, with the first time showing two images of hard grains and a soft gel material, while the second time showing the same question, two images of raw and cooked pasta are displayed.These illustrative examples demonstrate that the diagrams do not depict complex situations. Instead, by comparing them to objects familiar to the user from everyday life, the binary decision-making process for each question is significantly simplified, allowing for quick and intuitive answers. This approach also enables users who are not experts in bulk materials and their flow properties to easily select a suitable dosing device for their intended application and determine the appropriate operating parameters. This allows them to convey and dose their specified bulk material at the desired feed rate using the chosen dosing device, even without in-depth knowledge.

[0025] For this purpose, it may be provided, for example, that the opposing properties of the illustrations relate to physical characteristics of the bulk material that can be determined by the user without aids, such as granularity versus fineness, dryness versus moisture, or smooth versus rough surface.

[0026] Advantageously, the machine learning method optionally incorporates a neural network trained on a large number of bulk material samples with known flow properties. Neural networks can learn complex relationships between the various input information and flow properties, resulting in high predictive accuracy. The machine learning method can be continuously improved with new data to enhance its performance.The machine learning process can comprise different modules and, for example, largely independently of one another, determine the geometric parameters of the bulk material within the geometry evaluation module, based on optical characteristics determined from the optical recording of the bulk material. Simultaneously or subsequently, a separate module of the machine learning process can evaluate the user's binary decisions in order to then determine the relevant characteristic value, or potentially several characteristic values, of the flow properties in the characteristic value evaluation module, together with the geometric parameters. The machine learning process can be continuously improved by performing an increasing number of process runs for the same or different bulk materials and by subsequently evaluating the dosing device determined by the process and its operating parameters.

[0027] Many manufacturers of dosing devices have already used complex measurement methods to determine various parameters for the flow properties of a large number of bulk material samples and have also verified the suitability of individual dosing devices for the respective bulk material sample and an associated conveying rate range. The results of these measurement methods and verifications are usually stored in a database or can be transferred to a database. In order to utilize the measurements and verification results already performed for carrying out the method according to the invention, and in particular for the fastest possible learning process of the machine learning method, one embodiment of the invention provides for the use of a database containing a large number of bulk material samples for which the flow properties have been determined and measured through previous tests, and further comprises: the comparison of the geometric parameters of the bulk material provided by the user with the bulk material samples stored in the database; the selection of one or more bulk material samples from the database whose flow properties most closely match the determined geometric parameters of the bulk material; the use of the flow properties of the selected bulk material samples to determine the characteristic value for the flow properties of the bulk material.

[0028] In this way, the training of the machine learning model can be significantly accelerated and the results improved by using the previously painstakingly determined flow properties of several bulk material samples already stored in the database. This is achieved in the parameter evaluation module for determining the parameter based on the determined geometric parameters and the user's responses. By automatically comparing results obtained without comparison to the bulk material samples stored in the database and their determined flow properties with results obtained with this comparison, supervised learning and training of the machine learning model can be performed.In this way, the machine learning method in the key performance indicator (KPI) evaluation module can be continuously trained and improved very quickly and efficiently without additional effort.

[0029] The bulk material samples stored in the database can be used, for example, to determine the flow properties of the initially unknown bulk material by comparing the geometric parameters using a machine learning method. This method calculates the similarities between the geometric parameters of the bulk material provided by the user and the bulk material samples stored in the database. The comparison can be automated without manual intervention, thus accelerating the process. The machine learning method can improve the accuracy of the comparison by considering complex patterns and relationships.

[0030] It is also conceivable that the database contains information on a variety of physical properties of the bulk material samples, including particle size, density, moisture content and surface structure, and that this information is used to improve the accuracy in selecting the appropriate bulk material sample.

[0031] According to one embodiment of the invention, the bulk material samples stored in the database may be generated through experimental tests to determine the flow properties under various conditions, including different moisture levels, temperatures, and particle size distributions. Considering these different conditions makes the method more robust to varying environmental conditions. Furthermore, the user can be presented with and queried about environmental conditions or other properties of the bulk material in order to more precisely determine the characteristic value for the flow property by taking into account the various data sets experimentally determined for the same bulk material under different conditions.During the dialog, environmental conditions or other conditions can be presented as additional questions to elicit corresponding binary decisions from the user. It is also conceivable that the environmental conditions or other conditions, for example, could be entered by the user via the input module as real numerical values ​​within a defined range, just like the production rate.

[0032] The bulk material samples stored in the database can be classified into groups based on similar flow properties, and a dosing device selected from the most suitable group can be suggested to the user. Classifying the bulk material samples reduces the effort required to determine the appropriate dosing device for the calculated flow property parameters without significantly compromising the quality of the results.

[0033] Optionally, the database can also include simulation results that model the dosing device under various operating conditions in order to calculate the best combination of dosing device and operating parameters for the bulk material provided by the user. The simulation results can also be stored in and retrieved from a separate simulation database.

[0034] To ensure the most user-friendly process possible, the flow properties of the selected bulk material samples from the database are used in real time to refine the determined parameters for the flow properties of the bulk material provided by the user. In this way, the user can receive a suggestion for a suitable dosing device and its operating parameters for dosing the initially unknown bulk material immediately after completing their input. If the user subsequently changes any of the parameters, the effects of these changes can be displayed immediately.

[0035] Optionally, the comparison of the geometric parameters of the user-supplied bulk material with the flow properties of the bulk material samples stored in the database can be performed by a machine learning model that has learned the relationships between geometric parameters and flow properties. The comparison between the geometric parameters of a bulk material and the flow properties stored in the database can, for example, be based on an indirect approach. The geometric parameters, such as the volume, the height of the pile, or the angle of the heap, provide clues about the physical properties of the bulk material, such as particle size, density, or cohesion, which in turn influence the flow properties. Machine learning methods can be used to establish relationships between the geometric parameters and the physical flow properties.A suitable parameter for the flow properties of the bulk material can then be determined by comparison with the bulk material samples stored in the database.

[0036] To facilitate the most efficient and accurate comparison of the bulk material with the bulk material samples stored in the database, it is optionally possible for the machine learning algorithm to be trained to estimate the physical properties of the bulk material based on geometric parameters. These estimates are then compared with the physical properties stored in the database to find the best possible match. In this way, the machine learning algorithm integrates information from different sources to maximize prediction accuracy.

[0037] It is also conceivable that the geometric parameters are used to estimate specific flow properties of the bulk material, such as the bulk density, the angle of repose or the flow behavior, and that these flow properties are compared with the values ​​stored in the database for the bulk material samples.

[0038] According to an advantageous embodiment of the invention, the proposals for the dosing device and the operating parameters are calculated in real time based on the information entered by the user and the determined characteristic value. This real-time calculation enables immediate feedback to the user and rapid information about the suitable dosing device and its operating parameters. If the method takes into account data sets of bulk material samples stored in an external database, the electronic data processing device used to carry out the method must be able to establish a data transmission connection with the database and retrieve the necessary information from the database in real time.The query for the required information from the database can be limited to data records that are even relevant based on an initial evaluation of the information entered by the user, so that not all data records from the database have to be retrieved every time the procedure is carried out, or have to be stored in an internal storage unit of the electronic data processing device after previous retrievals.

[0039] In many cases, a mobile device can be a particularly suitable electronic data processing device for carrying out the procedure. Mobile devices can often produce photographic images of the bulk material with sufficient image quality. Most mobile devices, with their graphical display capabilities, available storage capacity, and processing power, are readily capable of handling the procedure in a user-friendly manner. Furthermore, most mobile devices have interfaces for standardized communication and data transmission, enabling a connection to a database and automated querying of database records without additional devices or components.

[0040] According to one embodiment of the invention, the operating parameters of the metering device are set within a predetermined range of values, which is determined based on the calculated characteristic value for the flow properties of the bulk material. This predetermined range ensures that the operating parameters are varied only within safe and effective limits. Furthermore, the predetermined range can be specified or limited depending on the specific metering device.

[0041] Optionally, the operating parameters can include continuous values ​​such as a conveying speed, a dosing frequency, an inclination angle, or vibration settings, and a real numerical value within the specified range can be proposed for each operating parameter. By not limiting the operating parameters to single or discrete values, the method can determine and propose the best possible operating parameters without, for example, imposing limitations on the application of the proposed operating parameters to the operation of the proposed dosing device, which would otherwise affect the result.If necessary, in addition to the operating parameters suggested as real numerical values, device-specific operating parameter settings can also be determined and displayed, taking into account the device-specific characteristics and limitations when entering or specifying the operating parameters. This simplifies the operation of the dosing device for the user and ensures reliable dosing.

[0042] Optionally, it can be provided that the specified value range for the operating parameters is calculated on the basis of bulk material samples with similar flow properties stored in the database.

[0043] According to one embodiment of the invention considered advantageous, the graphical user interface displays the suggested operating parameters for the dosing device in an adjustable slider or input field, allowing the user to manually adjust the values. These adjustments are visualized in real time. The graphical representation helps the user better understand the relationships between the flow properties and the dosing parameters. Furthermore, this visualization facilitates the selection of the optimal operating parameters and improves the efficiency of the process flow.

[0044] After determining the characteristic value, the graphical user interface can also display a graphical representation of the flow properties and their effects on the conveying and dosing processes to the user, in order to facilitate the selection and adjustment of the operating parameters as well as to verify the proposed dosing device and the operating parameters.

[0045] The data stored in the database can be used for automated selection of the dosing device by defining characteristic flow velocity ranges for each device and comparing them with the flow properties determined for the bulk material. The system can then optimize the operating parameters for the selected dosing device in real time by using a machine learning algorithm to calculate the optimal values ​​for conveying speed, dosing frequency, and other parameters based on the bulk material samples and their operating parameters stored in the database.If the data stored in the database also contains physical test results for a bulk material sample, which were carried out for different dosing devices and possibly under different conditions, a particularly precise selection of the suitable dosing device and operating parameters can be made possible by taking this information into account.

[0046] According to a particularly advantageous embodiment of the invention, the operating parameters determined from the calculated flow properties are automatically transmitted by the electronic data processing unit to a dosing device selected by the user and used for subsequent operation of the dosing device. In this way, the user-selected dosing device can be set up and prepared for operation with the bulk material without any additional user interaction, thus making the process very convenient. Furthermore, errors in the operation of the dosing device, which could be attributed, for example, to careless mistakes or a lack of expertise when entering the operating parameters, are eliminated.

[0047] Optionally, the process can also include a dosing device operation monitoring step, during which real-time data on flow rate and conveying capacity are continuously acquired. A monitoring module then checks whether the originally specified operating parameters need to be adjusted to maintain or improve dosing efficiency. This allows for monitoring the ongoing operation of the dosing device with the bulk material, enabling the assessment of the consequences of any changes to an operating parameter based on previously acquired information used to determine the parameters, as well as any additional data sets. Based on this, changes to the operating parameters can also be automated.Furthermore, the actual consequences of an automated change to the operating parameters can be recorded and used to train a machine learning process, which can then determine and specify future automated changes to the operating parameters.

[0048] Exemplary embodiments of the invention, which are schematically illustrated in the drawings, are explained in more detail below. They show: Fig. 1 a schematic flowchart for a process flow according to the invention of a method for selecting a metering device and determining operating parameters for metering a bulk material with unknown flow properties, Fig. 2 a schematic representation of a system for selecting a dosing device and specifying operating parameters for the use of the dosing device for dosing the bulk material, Fig. 3 a schematic representation of a question within a dialogue in which a user is prompted to enter a binary decision via a graphical user interface, and Fig. 4 An exemplary representation of several questions that are presented to the user in succession in the dialogue for answering with a binary decision.

[0049] A in Fig. 1. Exemplary procedure flow for a process for selecting a in Fig. The process begins, in a first step 3, with the schematically depicted dosing device 1 and the determination of operating parameters for dosing a bulk material 2 with unknown flow properties using the dosing device 1, by creating an optical image of a predetermined quantity of the bulk material 2 in a reference environment 4. The reference environment 4 comprises a flat reference surface 5 with a uniform color that differs from the surroundings and whose size is known. Reference markings 6 with a scale division, visible in the recorded optical image, are placed on the reference surface 5.

[0050] In a subsequent process step 7, a geometry evaluation module of an electronic data processing device, which is located in Fig. 2, exemplified as a mobile phone 8, geometric parameters of the bulk material 2 are determined based on the previously acquired optical image. In the case of a planar arrangement of individual particles of the bulk material, the geometric parameters can relate to the size, size distribution, and shape of the particles. In the case of a Fig. In the case of the accumulation of a quantity of particles not shown, the geometric parameters can also include, for example, the height and volume of the accumulation as well as an angle of the cone surface.

[0051] In a further process step 9, a user is guided through a dialogue with an input module of the electronic data processing device. The user must answer a number of questions, each by making a binary decision.

[0052] Subsequently, in a process step 10, a characteristic value for the flow properties of the bulk material 2 is determined using a characteristic value evaluation module of the electronic data processing device, employing a machine learning method and based on the determined geometric parameters and the user's responses.

[0053] In a subsequent process step 11, the user can enter a desired conveying rate for the bulk material 2 via the input module.

[0054] Subsequently, in a process step 12, the appropriate metering device 1 and suitable operating parameters for this metering device 1 for metering the bulk material 2 are proposed, which are based on the determined characteristic value for the flow properties of the bulk material 2 and the desired conveying rate.

[0055] Optionally, if the user of the proposed dosing device 1 and the proposed operating parameters for it agrees, in a subsequent process step 13 the previously determined operating parameters can be automatically transmitted to a user-selected and identified instance of the dosing device 1, so that the relevant instance of the dosing device 1 can be operated with the operating parameters to convey and dose the bulk material 2.

[0056] If necessary, a dosing device operation monitoring step 14 can be carried out during the operation of the dosing device 1, within which real-time data on the flow rate and conveying capacity are continuously recorded and a monitoring module is used to check whether the originally specified operating parameters need to be adjusted in order to maintain or improve the efficiency of the dosing.

[0057] During the process, a database containing numerous bulk material samples, for which flow properties have been determined and measured through previous experiments, can be accessed. These experimentally determined flow properties can then be used to improve the process. For this purpose, the process can include the following steps: The relevant data sets for bulk material samples are retrieved from a database 15 and made available to the electronic data processing system. The geometric parameters of the user-supplied bulk material 2, determined in process step 7, are compared with the corresponding information of the bulk material samples stored in database 15. Subsequently, one or more bulk material samples are selected from the database whose flow properties most closely match the determined geometric parameters of the bulk material 2.The flow properties of the selected bulk material samples identified in this way are taken into account to determine the characteristic value for the flow properties of the bulk material 2.

[0058] In Fig. Figure 3 schematically illustrates how, with the aid of a graphical user interface 16, for example from a mobile device 8, a tablet, or other electronic data processing device, a question is displayed to the user during the dialogue and their answer, transmitted as a binary decision, is recorded. In addition to a short question text 17, for example the question "What is the shape of the bulk material?", two Fig. The screen displays two images showing different materials. The user decides which of the two images best answers the question posed and submits this binary decision, for example, by tapping the corresponding image. Fig. or by a swiping motion, or by swiping in the direction of what he considers more appropriate Fig. .

[0059] In Fig. 4 is an example, from top to bottom, of a sequence of questions and their two corresponding counterparts. Fig. The sequence of questions is illustrated. Several questions may contain the same question text 17, but each is combined with different images. This allows the user to gather extensive information about the property of the bulk material 2 queried by question text 17 through multiple binary decisions. The sequence of questions illustrates the respective inputs or binary decisions that a user might have made while processing the dialog.

Claims

[1] Method for selecting a metering device (1) and determining operating parameters for metering a bulk material (2) with unknown flow properties, comprising the steps: - Creating an optical image of a given quantity of the bulk material (2) in a reference environment (4); - Determining geometric parameters of the bulk material (2) based on the optical imaging with a geometry evaluation module of an electronic data processing device; - Guiding a user through a dialogue with an input module of an electronic data processing device, whereby the user answers a number of questions by a binary decision; - Determining a characteristic value for the flow properties of the bulk material (2) using a characteristic value evaluation module of an electronic data processing device using a machine learning procedure, based on the determined geometric parameters and the user's responses; - User input of a desired delivery rate via the input module; - Proposing a suitable metering device (1) and suitable operating parameters for this metering device (1) for metering the bulk material (2), based on the determined characteristic value for the flow properties and the desired conveying rate. [2] Method according to claim 1, characterized by, that in addition to the desired conveying rate, a desired dosing accuracy is entered by the user via the input module, and that a suitable dosing device (1) and suitable operating parameters for this dosing device (1) for dosing the bulk material (2) are proposed based on the determined characteristic value for the flow properties and the desired conveying rate and the desired dosing accuracy. [3] Method according to claim 1 or claim 2, characterized by , that the questions of the dialogue relate to physical properties of the bulk material (2), such as particle size, moisture content or surface roughness. [4] Method according to any one of the preceding claims, characterized by, that the dialogue is carried out using a graphical user interface (16) and that the questions posed in the graphical user interface (16) show the user two images (18, 19) with opposite properties of the bulk material (2), and the user selects by a binary decision which image (18, 19) better describes the properties of the bulk material (2) he has provided. [5] Method according to claim 4, characterized by , that the opposite properties shown in the figures (18, 19) relate to physical characteristics of the bulk material (2) which can be determined by the user without aids, such as granularity versus fineness, dryness versus moisture, or smooth versus rough surface. [6] Method according to any one of the preceding claims, characterized bythat the machine learning method comprises a neural network that was trained using a large number of bulk material samples with known flow properties. [7] Method according to any one of the preceding claims, characterized by , that a database (15) containing a large number of bulk material samples is used, for which the flow properties have been determined and measured by previous experiments, and the procedure further comprises: - Comparison of the geometric parameters of the bulk material provided by the user (2) with the bulk material samples stored in the database (15); - Selection of one or more bulk material samples from the database (15) whose flow properties most closely match the determined geometric parameters of the bulk material (2); - Use of the flow properties of the selected bulk material samples to determine the parameter for the flow properties of the bulk material (2). [8] Method according to claim 7, characterized by , that the comparison of the geometric parameters is carried out using a machine learning method which determines the similarities between the geometric parameters of the bulk material provided by the user (2) and the bulk material samples stored in the database (15). [9] Method according to claim 7 or 8, characterized by , that the database (15) contains information on a variety of physical properties of the bulk material samples, including particle size, density, moisture content and surface structure, and that this information is used to improve the accuracy in selecting the appropriate bulk material sample. [10] Method according to any one of claims 7 to 9, characterized by, that the bulk material samples stored in the database (15) were created by experimental trials to determine the flow properties under different conditions, including different levels of moisture, temperature conditions and particle size distributions. [11] Method according to any one of claims 7 to 10, characterized by , that the flow properties of the selected bulk material samples from the database (15) are used in real time to refine the determined parameter for the flow properties of the bulk material (2) provided by the user. [12] Method according to any one of the preceding claims 7 to 11, characterized by , that the comparison of the geometric parameters of the bulk material provided by the user (2) with the flow properties of the bulk material samples stored in the database (15) is carried out by a machine learning method that has learned relationships between geometric parameters and flow properties. [13] Method according to claim 12, characterized by , that the machine learning method was trained to estimate physical properties of the bulk material (2) based on geometric parameters, which are then compared with the physical properties stored in the database (15) to find the best possible match. [14] Method according to any one of claims 7 to 13, characterized by , that the geometric parameters are used to estimate specific flow properties of the bulk material (2), such as the bulk density, the angle of repose or the flow behavior, and these flow properties are compared with the values ​​stored in the database (15). [15] Method according to any one of the preceding claims, characterized by , that the suggestions for the dosing device (1) and the operating parameters are calculated in real time based on the information entered by the user and the determined characteristic value. [16] Method according to any one of the preceding claims, characterized by , that the operating parameters of the dosing device (1) are set within a predetermined range of values, which is determined on the basis of the determined characteristic value for the flow properties of the bulk material (2). [17] Method according to claim 16, characterized by , that the operating parameters include continuous values ​​such as a conveying speed, a dosing frequency, an inclination angle or vibration settings, and that for each operating parameter a real numerical value within the specified range of values ​​is proposed. [18] Method according to any one of claims 4 to 17, characterized by, that the graphical user interface (16) presents the proposed operating parameters for the dosing device (1) in an adjustable slider or input field to allow the user to manually adjust the values, with the adjustments from the proposed values ​​being visualized in real time. [19] Method according to any one of claims 4 to 18, characterized by , that the graphical user interface (16) shows the user, after the determination of the characteristic value, a graphical representation of the flow properties and their effects on the conveying and dosing processes, in order to facilitate the selection and adjustment of the operating parameters. [20] Method according to any one of the preceding claims 7 to 19, characterized by, that the data stored in the database (15) are used for the automated selection of the dosing device (1) by storing characteristic flow velocity ranges for each dosing device (1), which are compared with the flow properties determined for the bulk material (2). [21] Method according to claim 20, characterized by , that the operating parameters for the selected dosing device (1) are optimized in real time by a machine learning method calculating the optimal values ​​for conveying speed, dosing frequency and other parameters based on the bulk material samples stored in the database (15) and their operating parameters. [22] Method according to any one of the preceding claims, characterized by, that the operating parameters determined from the determined characteristic value of the flow properties are automatically transmitted by the electronic data processing device to a dosing device (1) selected by the user and are used for subsequent operation of the dosing device (1). [23] Method according to any one of the preceding claims, characterized by , that the method also includes a dosing device operation monitoring step (14) in which real-time data on flow rate and delivery capacity are continuously recorded and a monitoring module is used to check whether the originally specified operating parameters need to be adjusted in order to maintain or improve the efficiency of the dosing.

Images