Method for determining status information relating to a belt grinder using a machine learning system

Abstract

A method for determining status information relating to a belt grinder (10) is proposed, wherein the belt grinder (10) has at least one grinding belt (18) for grinding a workpiece (14), the method comprising at least the following method steps: • providing measurement data relating to the belt grinder (10), • determining the status information from the provided measurement data by means of a machine learning system (58), wherein the machine learning system (58) is configured to determine the status information based on the provided measurement data. The invention also relates to a machine learning system (58), a method for teaching the machine learning system (58), a computer program, a computer-readable storage medium and a computer device (56), which are each provided and configured to perform the proposed method. The invention also relates to a belt grinder (10) for performing the method and a grinding pad (12) for use in a belt grinder (10).

Description

Technical Field

[0001] The invention relates to a method for determining status information relating to a belt grinder by means of a machine learning system. The invention further relates to the machine learning system and to a method for teaching the machine learning system. The invention also relates to a computer program, a computer-readable storage medium and a computer device, which are respectively provided and arranged to carry out the proposed method. The invention also relates to a belt grinder for carrying out the method and to a grinding pad for use in a belt grinder. Background Art

[0002] The invention is based on a belt grinder having at least one grinding belt for abrasive machining of workpieces. Such belt grinders are known from the prior art, for example from EP 2 576 137 A1. In abrasive machining of materials, in addition to the settings (e.g. process parameters) and the state (e.g. wear of the belt grinder) of the belt grinder, the properties and state of the grinding belt used for grinding also significantly influence the quality of the grinding result achieved.

[0003] DE 10 2017 120 260 A1 proposes a method for monitoring a grinding belt, wherein emissions emanating from the grinding belt during machining of a workpiece are determined, wherein a reference curve of the state of the grinding belt with respect to heating or with respect to sound waves emanating from the grinding belt is compared with actual values ​​of the heating and sound waves.

[0004] There is a need to be able to better control and monitor the influencing variables mentioned, in order to be able to achieve higher quality grinding results, less grinding belt wear and a longer service life of the belt grinder in this way. Summary of the invention

[0005] The invention is based on a belt grinding machine having at least one grinding belt for abrasive machining of a workpiece.

[0006] In a first aspect, the invention relates to a method, in particular a computer-implemented method, for determining status information relating to a belt grinding machine, wherein the method comprises at least the following method steps:

[0007] • provide measurement data concerning the belt grinder,

[0008] • The state information is determined from the provided measurement data by means of a machine learning system, wherein the machine learning system is configured to determine the state information based on the provided measurement data, in particular also output the state information.

[0009] In one embodiment of the method, the method for determining the status information can be implemented only in hardware. In an alternative embodiment, the method can also be implemented in the form of software or in the form of a mixture of software and hardware. In particular, the method can be a computer-implemented method executed by means of a computer device. In order to execute the method, the computer device can have at least one processor device and at least one storage device, in which the method is stored as a computer program.

[0010] A belt grinder is understood to be a tool for grinding workpieces, in which the grinding device is implemented in the form of a circumferential grinding belt. In the following, the term "belt grinder" includes grinding belts used in belt grinders and grinding pads used in belt grinders, etc. Belt grinders can be implemented, for example, in the form of hand-held belt grinders ("belt grinders") or in the form of large-scale belt grinding facilities on an industrial scale. Such industrial belt grinders are used, for example, for grinding workpieces over large areas, such as panels made of raw materials (such as particleboard, MDF panels, HDF panels and OSB panels). During the grinding process, the grinding belt moves relative to the surface of the workpiece to be processed, in particular continuously, so as to exert an abrasive action on the workpiece. Here, the grinding belt is typically driven by a roller in the direction of rotation of the belt and is pressed onto the surface of the workpiece to be processed in the direction of pressure by means of a grinding pad. In the case of an industrial belt grinder, the length of the grinding belt in the direction of rotation of the belt can be, for example, between 1.5 meters and 5 meters. In such a belt grinder, the width of the grinding belt, i.e. the width perpendicular to the direction of rotation of the belt, may be, for example, between 0.3 m and 5.0 m. In the case of a hand-held belt grinder, the length of the grinding belt in the direction of rotation of the belt may typically be between 0.35 m and 1 m. The width of the grinding belt of a hand-held belt grinder is typically between 7.5 cm and 15 cm. Other lengths and widths are also conceivable.

[0011] "Providing measurement data" is to be understood in particular as providing or transferring or inputting corresponding measurement data - relating to the belt grinder - to a device that executes the method, in particular a processor device or a computer device or a computer device that includes a processor device. In one embodiment of the method, the corresponding measurement data can be provided by reading in measurement data from a file. For example, the measurement data can be provided by reading in data stored on a data server or in a storage device stored on the belt grinder. In an alternative or additional embodiment of the method, the measurement data can be determined or measured using sensors, in particular sensors of the belt grinder, and then provided. In this case, a sensor is to be understood as a measuring technology device that can be used to detect measurement data relating to the belt grinder. In one embodiment of the method, at least one acoustic sensor, in particular an airborne sound sensor and / or a structure-borne sound sensor and / or a vibration sensor is used to measure and provide the measurement data. The acoustic sensor enables the detection of mechanical vibrations (sound waves) emanating from the belt grinder and / or the grinding belt. The measurement data detected and provided by means of an acoustic sensor relate to a belt grinder and in particular allow the mechanical properties of a grinding belt and / or a belt grinder to be analyzed and conclusions drawn during the grinding process. The mechanical properties can, for example, indicate defects in the grinding belt or clogging or wear of the grinding belt. Here, clogging refers to the filling of the gaps between the abrasive particles on the grinding belt by grinding debris or grinding dust. In addition, characteristics of the grinding belt, such as a targeted increase on the grinding belt (for example in the form of a barcode) can be read in and instructions about the grinding belt used can be derived therefrom. In particular, it is also conceivable to use an acoustic sensor that is specially set and / or limited with respect to the detectable frequency range. In particular, it is conceivable to use an infrasound sensor (typically below 16 Hz), an audible sound sensor (typically from 16 Hz to 20 kHz), an ultrasonic sensor (typically from 20 kHz to 1.6 GHz) and / or a hypersonic sensor (typically above 1 GHz) or a combination thereof. In one embodiment of the method, the at least one acoustic sensor and in particular the frequency range in which the at least one acoustic sensor can effectively detect is selected in accordance with the specific belt grinder. In particular, structure-borne sound sensors and vibration sensors make it possible to detect shocks, vibrations, resonances, etc., which are imperceptible or barely perceptible to humans, in the measurement data and thus provide them to the method according to the invention for determining status information relating to the belt grinder. In one embodiment of the method, the at least one acoustic sensor is implemented, for example, by a MEMS microphone sensor and / or by a laser microphone sensor and / or by a piezoelectric sensor or the like. Such sensors are known to the person skilled in the art. The evaluation of the measurement data detected and provided by means of the at least one acoustic sensor comprises in particular the finding of characteristic frequency components and / or sound amplitudes.In an alternative or additional embodiment of the method, at least one further sensor is used to detect (measure) and provide the measurement data, wherein the at least one further sensor is selected from a sensor list, the list comprising:

[0012] • Current consumption sensor for a belt grinder, in particular for a drive of a grinding belt of a belt grinder: For example, a coil can be provided as a sensor, with the aid of which the current consumption of the drive motor is detected inductively. The measurement data provided in this way relate to the belt grinder, so that they allow an analysis and conclusions to be drawn, in particular about the state of the grinding belt (blockage, wear, etc.) and / or the state of the belt grinder (bearing wear, drive motor wear, etc.). For example, it has been found that the current consumption is correlated with the friction between the grinding belt and the workpiece, wherein a blockage of the grinding belt has a significant influence on the friction;

[0013] • Air temperature sensor: The measurement data provided in this way relate to the belt grinder, so that they allow analysis and conclusions to be drawn, in particular regarding the heating during the grinding process, which in turn can be correlated with clogging or wear of the grinding belt, etc. and / or with wear of the belt grinder (bearing wear, drive motor wear, etc.);

[0014] • air humidity sensor and / or humidity sensor: the measurement data provided in this way relate to the belt grinder, so that they allow analyses and conclusions to be drawn, in particular regarding the material properties of the workpiece, which in turn are relevant to the grinding and / or cutting behavior of the grinding belt, the service life of the grinding belt, the surface quality of the grinding result achieved, clogging of the grinding belt (e.g. sticking at high humidity), etc.;

[0015] • distance sensors and / or displacement sensors, in particular laser distance sensors or laser displacement sensors: the measurement data provided in this way relate, for example, to a belt grinder, so that these measurement data allow analyses and conclusions to be drawn, in particular regarding vibrations of the grinding belt (“belt chatter”) and thus stresses (“tensile stresses”) in the grinding belt;

[0016] • Imaging, in particular imaging with a visual sensor, in particular a camera, a high-speed camera, a camera with a flash or strobe light, which operates, for example, at a wavelength in the visible spectrum or in the UV spectrum or in the infrared spectrum: the measurement data provided in this way relate to the belt grinder, so that they allow, in particular, an analysis and conclusions to be drawn about the optical properties of the grinding belt and / or the belt grinder during the grinding process. The optical properties can, for example, provide information about defects in the grinding belt or about blockages or wear of the grinding belt. Furthermore, features of the grinding belt, such as a barcode or QR code applied in a targeted manner, can be read in and information about the grinding belt used can thus be derived;

[0017] • Temperature sensor, in particular infrared sensor, in particular thermal imaging sensor: The measurement data provided in this way relate to the belt grinder, so that these measurement data allow, in particular, analyses and conclusions to be drawn about the temperature and / or the temperature distribution of the grinding belt and / or the belt grinder during the grinding process. For example, the thermal properties can provide information about defects in the grinding belt or about clogging or wear of the grinding belt. For example, clogging of the grinding belt is associated with an increase in friction and thus leads to an increase in the temperature of the grinding belt. Furthermore, the measurement data can relate to the temperature and / or the temperature distribution of the workpiece (for example before, during and after the grinding process) and likewise enable analyses and conclusions to be drawn about the belt grinder;

[0018] • thickness measuring sensor: the measurement data provided in this way relate to the belt grinder, so that they allow analysis and conclusions to be drawn, in particular regarding material removal from the workpiece and / or clogging of the grinding belt, etc., and thus, for example, regarding wear of the grinding belt;

[0019] • Torque sensor: The measurement data provided by means of the torque sensor relate to the belt grinder, so that they allow, in particular, the drive of the grinding belt to be analyzed and conclusions to be drawn and, for example, can be correlated with the friction between the grinding belt and the workpiece, which friction in turn can be influenced, for example, by a blockage of the grinding belt;

[0020] • Dust quantity measuring sensor: The measurement data provided in this way relate to the belt grinder, so that they allow, in particular, analyses and conclusions to be drawn regarding material removal from the workpiece and / or clogging of the grinding belt, etc. and thus regarding wear of the grinding belt. The dust quantity measuring sensor can be arranged, for example, in the exhaust gas duct of the belt grinder and detects the amount and / or the particle size distribution of the grinding dust generated;

[0021] • Acceleration sensors, travel distance sensors, inertial sensors: The measurement data provided in this way relate to the belt grinder, so that they allow, in particular, an analysis and conclusions to be drawn about the mechanical movements of the grinding belt and / or the belt grinder. For example, the measurement data can relate to components within the belt grinder or to the workpiece (e.g. feed speed during the grinding process). The measurement data relate in particular to deviations from operating parameters and / or desired operating parameters of the belt grinder and can therefore also make statements about the wear of the grinding belt and / or the belt grinder;

[0022] • Positioning sensors: The measurement data provided in this way relate to the belt grinder in such a way that they allow analyses and conclusions to be drawn, in particular about objects introduced into the belt grinder and / or the workpiece, such as metal objects (e.g. nails) and / or about objects introduced into the grinding belt during the grinding process, such as metal splinters. The measurement data can thus allow explanations of possible (foreseeable) defects in the grinding belt or possible hazards during the operation of the belt grinder. Exemplary positioning sensors may include magnetic field sensors, radar sensors, inductive sensors, capacitive sensors, nuclear magnetic resonance sensors, etc.;

[0023] • Touch-sensitive sensor: The measurement data provided by means of the touch-sensitive sensor relate to the belt grinder, so that these measurement data allow, in particular, the tactile surface properties of the workpiece and / or the grinding belt to be analyzed and conclusions to be drawn (e.g. roughness) and can be related, for example, to the friction between the grinding belt and the workpiece. Furthermore, the touch-sensitive sensor can be used to determine "belt chatter" (vibrations on the grinding belt);

[0024] • Reflection sensor: The measurement data provided by means of the reflection sensor relate to the belt grinder so that they allow, in particular, the surface properties (e.g. roughness) of the workpiece and / or the grinding belt to be analyzed and conclusions to be drawn and can, for example, be correlated with the friction between the grinding belt and the workpiece. Typical reflection sensors can be implemented, for example, using laser radiation, which is reflected on the surface to be checked and the intensity of the reflected laser light is detected;

[0025] or a combination of these sensors. The measurement data of the sensors from the list are very suitable for determining the status information according to the method according to the invention. By providing measurement data detected by means of a plurality of, in particular different sensors, the status information relating to the belt grinder can be determined particularly accurately and reliably. In particular, measurement data adapted to the belt grinder to be analyzed can be provided in this way for carrying out the method. In addition, partially redundant measurement data make it possible to avoid erroneous analyses, thereby making it possible to determine the status information more accurately. In addition, different and / or complementary and / or redundant status information can be determined based on the measurement data detected by means of a plurality of, in particular different sensors.

[0026] In one embodiment of the method for determining status information, the measurement data are selectively or selectively provided or called, in particular selectively or selectively provided or called from the belt grinder. For example, it can be provided that the measurement data are stored on a storage device, in particular a network storage device (server) or a storage device of the belt grinder, wherein the measurement data are selectively called from the storage device and then provided to a computer device that executes the method, such as a server in the cloud. It can thus be achieved in particular that the amount of data to be transmitted for executing the method is variable and can be selectively provided in particular depending on the status information to be determined. This means that only those measurement data that are actually required for executing the method for determining specific status information are provided, and other measurement data that may be present are not provided. In one embodiment, the measurement data are prepared in a database or as in a database. In particular, the status information that is of interest to the user can be selected by an input or selection of the user (for example by means of an input device or by means of a menu selection, etc.).

[0027] In one embodiment of the method for determining state information, the measurement data are filtered before being provided. In particular, speech components (in particular human speech) are filtered out of the measurement data. Alternatively or additionally, other interfering influences, such as background signal components, are filtered out of the measurement data. In this way, interfering influences in the measurement data that impair the reliable determination of the state information can be removed. In particular, interfering influences for which the machine learning system has not been trained and / or for which the machine learning system can only be trained with great effort are removed from the measurement data. According to current knowledge, speech components in the measurement signal represent particularly large interfering influences.

[0028] Status information relating to a belt grinder is understood to mean information which allows the description of the status of the belt grinder (e.g. progression of wear, necessity of maintenance) and / or the status of the grinding belt (e.g. progression of wear, necessity of replacement) and / or the operation of the belt grinder (e.g. process parameters, recommendations for changing process parameters). In particular, the determination of the status information makes it possible to increase the machine efficiency, the process efficiency and / or the operating efficiency, for example by early indication of problems, errors, necessary activities (e.g. maintenance) and the resulting reduction in machine downtimes, reduction in rejects (maintaining high quality during the grinding process) and reduction in the training requirements for users operating the belt grinder. In one embodiment of the method, the status information is defined or selected such that the status information relates to at least one of the following properties:

[0029] • Characterize the properties of the workpiece to be machined, such as material type such as "wood", "metal" or material properties such as "high hardness", "medium hardness", etc.;

[0030] • Characterize the properties of manufacturing defects on the workpiece, such as so-called dynamic striations, “chatter marks” or roughness on the workpiece;

[0031] • properties that characterize the operating mode or operating parameters of the belt grinder, such as power consumption, feed rate of the workpiece or "summer operation" / "winter operation" (the season has an influence, in particular, on the humidity of the grinding belt, which in turn influences the elasticity of the grinding belt and thus the removal rate, service life, etc. of the grinding belt);

[0032] •Attributes that characterize incorrect settings of the belt grinder, such as excessive grinding belt stress or excessive grinding belt rotation speed or inappropriate grinding pad pressure;

[0033] • properties that characterize the load distribution of the belt grinder, such as uneven wear of the grinding belt over the entire grinding belt width and / or uneven wear of the individual grinding belts when multiple grinding belts are used in parallel;

[0034] • Characterization of the degree of wear or the properties of wear in belt grinders, e.g. aging processes of bearings, rollers, grinding pads, wear of machine components, parameter drift, etc.;

[0035] • Characterizing the properties of the grinding belt used in the belt grinder, for example a grinding belt type such as "P120" or preferably used settings for optimizing the use of the grinding belt (grinding belt revolutions, grinding belt stress, etc.);

[0036] • Characterize the properties of grinding belt clogging and / or dulling,

[0037] • properties characterizing grinding belt defects, such as the presence of cracks in the grinding belt or the presence of foreign particles such as metal fragments in or on the grinding belt;

[0038] Or a combination of these.

[0039] In one embodiment of the method, it is also conceivable that different status information is determined or that the status information to be determined can be selected by a user of the belt grinder. In this way, a particularly comprehensive and flexibly adaptable method can be described, which can be used by the user to obtain a particularly clear overview of the status of the belt grinder, the grinding belt and / or the workpiece.

[0040] The state information is determined from the provided measurement data by means of a machine learning system, wherein the machine learning system is configured to determine the corresponding state information based on the provided measurement data, and in particular also output the state information. The machine learning system should be understood here in particular as a technical implementation of a self-learning system, which learns from a predetermined example (so-called training data) and can generalize the learned behavior after the learning phase is over by identifying templates and laws in the training data and making them callable. Such machine learning systems are known in principle, for example from DE 102005 050 577 A1. It is proposed that in one embodiment of the method for determining the state information, the machine learning system includes a neural network, in particular an artificial neural network. The neural network is composed of a link of neural layers, wherein the topology of the network is configured and adapted, in particular parameterized, to perform the method. Here, the measurement data is first provided to the machine learning system, in particular the neural network, as an input variable (training input data), and then the training input data is propagated through the machine learning system, in particular through the neural network. In particular, in the case of a neural network, each (hidden) layer of the neural network correspondingly calculates an output variable, which is in turn used as an input variable for the subsequent layers of the network. The last layer (output layer) of the network allows reading out the corresponding state information, which is estimated based on the input measurement data. A machine learning system, in particular a neural network such as a Bayesian network, has the following advantage here: the state information relating to the belt grinder (including the grinding belt and the workpiece) can be determined more reliably and more accurately than existing statistical schemes (for example, known from the prior art in DE 102017 120 260 A1). In particular, even with a large amount of measurement data and different influencing factors on the state information, meaningful results can be obtained when determining the corresponding state information. In one embodiment of the method, the neural network is implemented as a recursive neural network or a convolutional neural network. It is also conceivable that the machine learning system performs a regression, i.e. predicts the course of the state information.

[0041] In one embodiment of the method, the belt grinder is controlled at least partially based on the determined state information and / or information is outputted at least partially based on the determined state information by means of an output device. In order to control the belt grinder, the determined state information is output, in particular transmitted, to a control device of the belt grinder. Thus, by means of the control device, control variables for actuating a physical actuator of the belt grinder can be determined using the state information. The control device is used here to control, in particular to operate the physical actuator, for example by applying a regulation routine and / or a control routine. The control device is at least configured to perform further processing at least partially based on the determined state information and in this way convert the corresponding state information into control variables for actuating the actuator. For example, it is conceivable to automatically set process parameters, such as the belt rotation speed, as a function of the determined state information. In this way, a particularly effective monitoring method can be described for monitoring and controlling a belt grinder, in particular for automatically adapting a belt grinder. In this way, a belt grinder can be functionalized as an at least partially autonomous belt grinder. Thus, in particular, the belt grinder can automatically start the grinding process, automatically end the grinding process and / or automatically select and / or influence at least one parameter related to the execution of the grinding process. The output device can be an output device of a computer device that executes the method and / or an output device of the belt grinder and / or a separately constructed output device. For example, the separately constructed output device can be realized by a computer, a mobile data device such as a tablet computer, etc. It is conceivable, for example, to output information using a tactile, auditory or visual output device at least partially based on the determined state information. In particular, a screen can be used for output in graphical form. Alternatively or additionally, a data communication device can be used to output to an external device. It should be noted that using the method for determining state information according to the invention, in particular in combination with the output of information at least partially based on the determined state information, the determined state information related to the belt grinder can advantageously be perceived by a human user. The information here refers to information prepared to be output by means of an output device, in particular information prepared in a user-friendly manner and at least partially based on the determined state information. In particular, the information for output can also correspond to the state information. For example, it is conceivable that information in the form of a traffic light signal, which informs the user of the belt grinder about the time of changing the grinding belt, is output directly at the belt grinder.

[0042] According to another aspect of the invention, a method for teaching a machine learning system, in particular a neural network, is provided, in particular a computer-implemented method. The method may be processor-assisted. The method for teaching results in the machine learning system being configured, i.e. specifically taught and / or parameterized, to perform the above-described method for determining state information relating to a belt grinder. The method for teaching has at least the following method steps:

[0043] • providing training data consisting of training input data and training output data, wherein the training input data include measurement data relating to a plurality of status information relating to the belt grinder, and the training output data each include at least one assigned status information relating to the belt grinder,

[0044] • Teaching a machine learning system, in particular a neural network, wherein parameters of the machine learning system are adapted such that the machine learning system determines respectively associated training output data as a function of the adapted parameters and as a function of provided training input data.

[0045] In a further method step, the taught machine learning system is added to a computer device, in particular a control device of a belt grinder to be monitored, ie a belt grinding machine.

[0046] As a result of providing the training input data, the machine learning system, in particular the neural network, determines a first output value from the data. During the teaching process, the output value is fed to a training system (e.g., a computer device), wherein the training system determines therefrom a rule for adapting parameters, which predetermines which parameters of the machine learning system should be adapted in what manner to more accurately determine the predetermined training output data. In the case of a neural network, such adaptation can be performed by predetermining an expected value or expected value of the output value and subsequent back propagation. It is also proposed that in one embodiment of the method for teaching, the training input data is selected from a list, the list comprising at least measurement data provided by sensors of the following list: sensors for current consumption, air temperature sensors, air humidity sensors and / or humidity sensors, distance sensors and / or displacement sensors, imaging (in particular visual) sensors, temperature sensors (in particular infrared sensors, in particular thermal image sensors), thickness sensors, torque sensors, dust quantity sensors, acceleration sensors, travel distance sensors, inertial sensors, positioning sensors, touch sensors, reflection sensors, or combinations of these sensors. It is also proposed to select the training output data from a list of status information relating to at least the following properties: a property characterizing a workpiece to be machined; a property characterizing a manufacturing defect on the workpiece; a property characterizing an operating mode or operating parameter of a belt grinder; a property characterizing an incorrect setting of a belt grinder; a property characterizing a load distribution of a belt grinder; a property characterizing a degree of wear or a property characterizing a wear of a belt grinder; a property characterizing a grinding belt used in a belt grinder; a property characterizing a clogging and / or dulling of a grinding belt; a property characterizing a defect of a grinding belt or a combination of these properties. In this way, a particularly targeted method for teaching can be described, which is adapted to the provided measurement data to be advantageously used in a method for determining status information relating to a belt grinder, since it uses the same information source as a basis for teaching.

[0047] In one embodiment of the method for teaching, in a further method step, further measurement data relating to the belt grinder, in particular to be monitored, are provided, wherein at least one, in particular predetermined, state information relating to the belt grinder is assigned to the further measurement data, in particular by an expert, and in this way the machine learning system is further taught with the further measurement data. “Further teaching” is to be understood in particular as repeated teaching of the machine learning system with the now provided further training data, in particular of the belt grinder to be monitored. In this way, it is possible to first pre-train the machine learning system with the “generic measurement data”, then assign the machine learning system to a computer device, in particular a control device, in particular a control device of the belt grinder to be monitored, and then further train or further teach the machine learning system with the training data coordinated with the belt grinder to be monitored. As a result, a machine learning system that is particularly well trained on the respective belt grinder can be achieved, which is also able to particularly reliably determine the state information relating to the respective belt grinder.

[0048] In one embodiment of the method for teaching, the training input data include measurement data and further measurement data relating to a plurality of state information, and the training output data each include at least one assigned state information, wherein the measurement data and the further measurement data relate to

[0049] • At least two different types of belt grinders, or

[0050] • at least two belt grinders of the same type with different applications (i.e. with grinding processes that differ from one another, in particular with regard to the processed materials, process parameters, etc.), or

[0051] •Two belt grinders of the same type with the same purpose.

[0052] In particular, the first belt grinder is a belt grinder for which generic training data is provided. In particular, the second belt grinder is a belt grinder to be monitored, to whose computer device, in particular its control device, the machine learning system has been added. Thus, the first belt grinder can be identical to the second belt grinder, or alternatively, the first belt grinder can also be different from the second belt grinder. In this case, the second belt grinder is a further belt grinder. The first of the three alternatives makes it possible to illustrate a particularly generic machine learning system that can determine status information for a large number of belt grinders. The second alternative makes it possible to illustrate a machine learning system that is particularly suitable for belt grinders, which can determine status information related to the belt grinder for a large number of grinding processes (in particular using a large number of materials, process parameters, etc.). The third alternative makes it possible to illustrate a machine learning system that is particularly dedicated to belt grinders, which can determine status information related to the belt grinder for a relatively small number of grinding processes (in particular using a large number of materials, process parameters, etc.). However, a particularly high level of reliability and robustness in determining the status information can be achieved here.

[0053] In another aspect of the invention, a machine learning system, in particular a neural network, is proposed for executing the method according to the invention for determining state information. The machine learning system is obtained in particular by executing the method according to the invention for teaching a machine learning system.

[0054] In another aspect, a computer program is provided. The computer program is configured to perform one of the above methods. The computer program comprises instructions which, when executed on a computer device, i.e. in particular when executed by a processor device of the computer device, cause the execution of one of the described methods and all its steps.

[0055] Furthermore, a computer-readable storage medium is proposed, on which the computer program is stored, in particular stored. Storage media are known per se to those skilled in the art.

[0056] Furthermore, a computer device, in particular a control device for a belt grinder, is proposed, which is configured to carry out one of the methods described. In particular, a computer device for determining status information relating to a belt grinder is proposed, which has at least one processor device and a storage device, wherein instructions are stored on the storage device, which instructions, when executed by the processor device, cause the computer device to carry out the method. A "processor device" is to be understood in particular as a device having at least one information input, an information processing unit for processing, and an information output for forwarding processed and / or evaluated information. In one embodiment, the processor device comprises at least one processor. A "storage device" is used here to prepare a computer program required for the processor device to carry out one of the methods described.

[0057] In one embodiment of the computer device, the computer device is configured to use the determined state information to determine control variables for manipulating a physical actuator, in particular a control variable for manipulating a belt grinder, and / or a control variable for manipulating an output device, and / or to perform a function. In one embodiment, the physical actuator may be implemented as an automatic goods storage system, by means of which the required grinding belts, grinding pads, etc. are automatically output. For example, the function may be implemented by a system setting or a process setting of the belt grinder that is changed according to the determined state information.

[0058] In another aspect of the invention, a belt grinder is proposed, comprising a grinding belt for abrasive machining of a workpiece, characterized by at least one acoustic sensor for detecting and providing measurement data for carrying out one of the methods according to the invention. In one embodiment of the belt grinder, at least one acoustic sensor, in particular an airborne sound sensor and / or a structure-borne sound sensor and / or a vibration sensor, is arranged on or in a grinding pad of the belt grinder. The grinding pad typically consists of a carrier device and a grinding pad sleeve, wherein the grinding pad sleeve is typically pushed into the carrier device or arranged on / in the carrier device in another way. The grinding pad including the grinding pad sleeve is provided for supporting the grinding belt moving relative to the grinding pad sleeve in the direction of rotation of the belt with low friction, while pressing the grinding belt against the workpiece (or acting as a resistance to the workpiece pressed against the grinding pad). The at least one acoustic sensor is preferably arranged or integrated on or in the carrier device of the grinding pad. Alternatively or additionally, the at least one acoustic sensor is arranged on or in the grinding pad sleeve. In this way, it can be achieved that the acoustic sensor is located particularly close to the grinding process and that the measurement data can be detected particularly effectively and in particular with little interference. In particular, by such an arrangement of at least one acoustic sensor, state information characterizing properties of the grinding belt and / or properties of the grinding process can be determined in a particularly reliable manner. In one embodiment, at least one acoustic sensor is located substantially centrally on the grinding pad and / or in the belt grinder - with respect to the grinding belt width perpendicular to the direction of rotation of the belt and / or with respect to the grinding pad width perpendicular to the direction of rotation of the belt. In this way, it can be achieved that particularly good signal detection can be achieved in the belt grinder. Alternatively or additionally, a plurality of acoustic sensors can be distributed over the width of the grinding belt and / or over the width of the grinding pad and / or can be arranged on the grinding pad and / or in the belt grinder in a manner distributed over the width of the grinding pad.

[0059] In an alternative or additional embodiment of the belt grinder, at least one acoustic sensor, in particular a further acoustic sensor, in particular an airborne sound sensor and / or a structure-borne sound sensor and / or a vibration sensor, is arranged in such a manner as to be assigned to a roller of the belt grinder. The roller can be selected as a contact roller, a deflection roller, a tension roller, a drive roller or the like. "Arranged in such a manner as to be assigned to a roller" means that the acoustic sensor is arranged in or on the roller or directly on the roller. For example, at least one acoustic sensor can be arranged on a roller suspension of a roller of the belt grinder. Alternatively, at least one acoustic sensor can also be integrated into the roller. It has been found that in this way it is possible to arrange the acoustic sensor particularly close to the location where the grinding belt feed occurs and to detect measurement data particularly effectively and in particular with low interference. In particular, by such an arrangement of at least one acoustic sensor, state information characterizing properties of the belt grinder and / or properties of the grinding process can be determined particularly reliably.

[0060] In one embodiment of the belt grinder, at least two acoustic sensors are arranged on both sides with respect to the width of the grinding belt and / or with respect to the width of the grinding pad on the grinding pad and / or in the belt grinder and / or assigned to the roller. Surprisingly, the arrangement of two acoustic sensors on both sides, in particular on the roller, has proven to be advantageous with regard to signal detection in a belt grinder.

[0061] In one embodiment of the belt grinder, at least one acoustic sensor is connected or connectable to the computer device, in particular a control device of the belt grinder and / or to a computer device outside the belt grinder, in particular using a gateway. The connection can be realized wired or wirelessly. For example, the connection can be realized using an Ethernet connection, an optical fiber connection, an Internet connection, a radio connection or a direct connection. In particular, the acoustic sensor is also connected or connectable via the gateway to other signal technology components of the belt grinder, such as further sensors.

[0062] In one embodiment of the belt grinder, the acoustic sensor is realized by a MEMS microphone sensor and / or by a piezoelectric sensor and / or by a laser microphone sensor.

[0063] In another aspect of the invention, a grinding pad for use in a belt grinder is proposed, comprising at least one acoustic sensor, in particular an airborne sound sensor and / or a structure-borne sound sensor and / or a vibration sensor, for providing measurement data for carrying out one of the proposed methods. As described above, the at least one acoustic sensor is in particular arranged in a carrier device of the grinding pad. Furthermore, the at least one acoustic sensor can be arranged substantially centered with respect to the width of the grinding pad, i.e. perpendicularly to the direction of rotation of the grinding belt. Alternatively, a plurality of acoustic sensors can be arranged on or in the grinding pad in a manner distributed over the width of the grinding pad.

[0064] In the following, "provided" and "provided" are to be understood exclusively as "programmed", "designed", "conceptualized", "parameterized" and / or "equipped". An object "provided" for a certain function is to be understood in particular to mean that the object fulfills and / or performs this certain function in at least one application state and / or operating state or is designed to fulfill this function.

[0065] It should be noted that features discussed in the description of the method that imply structural implementations can also be transferred analogously to corresponding devices that are also considered to be disclosed. Thus, for example, the frequency range of an acoustic sensor disclosed in conjunction with the method for determining status information can be directly transferred to an acoustic sensor operating in a corresponding frequency range. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In the following description, the invention is explained in more detail based on the embodiments shown in the drawings. The drawings, the specification and the claims contain many features in combination. Those skilled in the art will also appropriately consider these features individually and combine them into meaningful other combinations. The same reference numerals in the figures represent the same elements.

[0067] Figure 1 A schematic diagram of an embodiment of a belt grinder is shown in cross-sectional view,

[0068] Figure 2 A schematic diagram showing an embodiment of a grinding pad and a grinding belt is shown in a perspective view,

[0069] Figure 3 A schematic diagram showing an implementation of the new neural network,

[0070] Figure 4 A schematic diagram illustrating an embodiment of a method for teaching a machine learning system,

[0071] Figure 5 A schematic diagram shows an embodiment of a method for determining status information. DETAILED DESCRIPTION

[0072] Figure 1 A schematic diagram of an embodiment of a belt grinder 10 is shown in cross section, which has a grinding pad 12. During the grinding process, a workpiece 14 is ground on a grinding table 16 beneath a rotating grinding belt 18. The grinding belt 18 is driven by three rollers 20 (here drive rollers) in the direction of belt rotation 22 and is pressed against the workpiece 14 by the grinding pad 12.

[0073] Figure 2 The grinding pad 12 is shown in an enlarged perspective view. The grinding pad 12 comprises a carrier device 26 and a grinding pad cover 28. The grinding pad cover 28 has a cover carrier 30 made of MDF (alternatively also plastic or cardboard or fiber or metal material), to which the cushioning layer is bonded as a support cover 32 made of solid foam. A sliding cover 34 is bonded to the support cover 32, which is realized by a graphite-coated fabric. The carrier device 26 has a recess 36, into which the cover carrier 30 of the grinding pad cover 28 can be inserted in an insertion direction 38. The cover carrier 30 has a dovetail-shaped cross-sectional profile corresponding to the recess 36. The cover carrier 30 has an elongated shape with a length of 3000 mm in the insertion direction 38, wherein the length here extends in the direction of the width 40 of the grinding belt 18. The width of the cover carrier 30 perpendicular to the insertion direction 38 and perpendicular to the extrusion direction 42 has a width of 75 mm.

[0074] The belt grinder 10 comprises four acoustic sensors 44 for detecting and providing measurement data relating to the belt grinder 10 , three of which are structure-borne sound sensors 44 a, b, c (in Figure 1 The first structure-borne sound sensor 44a is screwed onto the surface 46 of the carrier device 26 of the grinding pad 12 facing the grinding belt 18 and detects the structure-borne sound there, which is transmitted by the grinding pad 12. In this way, the first structure-borne sound sensor 44a is arranged in the immediate vicinity of the grinding belt 18. A second structure-borne sound sensor 44b and a third structure-borne sound sensor 44c are screwed onto both sides of the roller 20 of the belt grinder 10 (see FIG. 2 ). Figure 1 ), so that two structure-borne sound sensors 44b, c are arranged on the roller 20 on both sides of the width 40 of the grinding belt 18. The structure-borne sound sensors 44a, b, c are implemented as structure-borne sound sensors ("AE-Sensor-S") from the company Dittel / Marposs, which detect acoustic signals in the frequency range of 250-300 kHz. The airborne sound sensor 44d is arranged in the center of the belt grinder 10, here fixed to a frame element of the belt grinder 10, which is not shown in detail. The airborne sound sensor 44d is obtained here, for example, from the company Mars Sensor and is a silicon MEMS microphone sensor. The detected frequency range is 55 Hz to 20 kHz. A speech analyzer (not shown in more detail here) is used to filter out speech components from the measurement data provided by the airborne sound sensor 44d.

[0075] Furthermore, further sensors 50 are provided in the belt grinder for detecting further measurement data relating to the belt grinder 10. The further sensors 50 include a sensor for power consumption (not shown in greater detail here) and two thermal image sensors 52, which are each aligned with the inner side of the surrounding grinding belt 18. The thermal image sensor 52 is located in front of the grinding pad 12, while the thermal image sensor 52 is located behind the grinding pad 12, as seen in the direction of rotation 22 of the grinding belt. The acoustic sensor 44 and the further sensors 50 are connected to a control device 54 of the belt grinder 10 by using a gateway 48 and further to an external computer device 56. The connection is wireless here, as indicated by a small radio symbol (three lines). With the aid of the acoustic sensor 44 and the further sensors 50, the measurement data are detected and forwarded to the control device 54, where they are stored in a storage device not shown in greater detail here. During the execution of the method for determining status information, the measurement data can be selectively and optionally called up from the storage device by the computer device 56 executing the method.

[0076] The computer device 56 provided for executing the method for determining status information relating to the belt grinder 10 is implemented as a server separate from the belt grinder 10. In another embodiment, the computer device 56 can also be integrated in the control device 54 of the belt grinder 10 or implemented by the control device. The computer device 56 is used to determine status information relating to the belt grinder 10. To this end, the computer device 56 executes a computer-implemented method (see Figure 5 ), the method comprises the method steps of providing measurement data relating to the belt grinder 10 and determining status information from the provided measurement data by means of a machine learning system 58. In this exemplary embodiment, the method makes it possible to determine status information that can be selected or predefined by a user of the belt grinder 10, see Figure 5 . For this purpose, the user can select the desired status information by means of an input device (here implemented as an input and output device 60 of the belt grinder 10) which can be connected to the computer device 56. Here, the user can select between nine status information, each of which relates to different aspects and properties of the belt grinder 10, the grinding belt 18 used and / or the grinding process. By means of the input and output device 60, the user can view or query the evaluation results, i.e. the determined status information. In addition, the computer device 56 is connected or connectable to the control device 54 of the belt grinder 10 via a data communication device (radio connection), so that the control variables determined using the status information for actuating the physical actuator (here, for example, the drive motor of the roller 20) can be output to the control device 54 and can thus be directly converted into activities on the belt grinder 10. The belt grinder 10 can thus be controlled at least partially based on the determined status information. In addition, information can be output to the user at least partially based on the determined status information by means of the input and output device 60 of the belt grinder 10.

[0077] When executing the method 200 for determining status information relating to the belt grinder 10, the computer device 56 implements a machine learning system 58, which is configured to determine status information based on the provided measurement data. In particular, the acoustic sensor 44 and the further sensor 50 are connected or can be connected to the computer device 56 via signal technology for this purpose. In this way, the provided measurement data are provided as input variables to the machine learning system 58. The machine learning system 58 then determines output variables, in particular corresponding status information relating to the belt grinder 10, as a function of a plurality of parameters of the machine learning system 58 (as mentioned at the outset, the term belt grinder 10 here also includes the components of the belt grinder 10 involved during the grinding process, in particular the grinding belt 18, the workpiece 14).

[0078] Figure 3A schematic diagram of a machine learning system 58 is shown, which in this embodiment is provided by a neural network 58a. The neural network 58a comprises a plurality of layers 62, each of which is linked to one another by means of connections 64 and each of which comprises a plurality of neurons 66. At least measurement data are provided as input variables 68 to the neural network 58a, which are then propagated through the neural network 58a. The neural network 58a determines output variables 70 layer by layer as a function of the input variables 68. To this end, each layer 62 determines an output variable 70 as a function of the input variables 68 provided to it and as a function of the parameters of the layer. The output variable 70 is then forwarded to the further layers 62 via the connections 64. The last layer 62a of the network 58a allows the reading of status information calculated based on the provided measurement data.

[0079] Figure 4A method diagram of an exemplary embodiment of a computer-implemented method 100 for processor-assisted teaching of a machine learning system 58, in particular a neural network 58a, is shown. The method 100 is performed by a training system (not shown in greater detail here) for teaching the machine learning system 58. In this case, training data are provided to the machine learning system 58 in a method step 102. The training data include training input data and training output data, wherein the training input data include measurement data for a large number of status information related to the belt grinder 10, and the training output data each include at least one assigned status information related to the belt grinder 10. In this exemplary embodiment, the training input data are measurement data from the acoustic sensor 44 and the further sensor 50. In this exemplary embodiment, the training output data relate to nine pieces of status information, each of which relates to a different property of the belt grinder 10, the grinding belt 18 used, the workpiece 14 and / or the grinding process: a property characterizing the workpiece 14 to be machined; a property characterizing a manufacturing defect on the workpiece 14; a property characterizing an operating mode or operating parameters of the belt grinder 10; a property characterizing an incorrect setting of the belt grinder 10; a property characterizing a load distribution of the belt grinder 10; a property characterizing a degree of wear or wear of the belt grinder 10; a property characterizing a grinding belt 18 used in the belt grinder 10; a property characterizing a clogging and / or dulling of the grinding belt 18; a property characterizing a defect of the grinding belt 18. The machine learning system 58, in particular the neural network 58a, is taught in method step 104. When teaching the machine learning system 58, the parameters of the corresponding layer 62 are adapted so that the machine learning system 58 determines the respectively assigned training output data from the provided training input data. The teaching of the machine learning system 58 can be carried out here using a difference function (cost function), which in particular characterizes the difference between the calculated output variable 70 and the training output data, wherein the difference function is optimized in terms of parameters by means of a gradient descent method. Such gradient descent methods are known to those skilled in the art from the prior art. After the parameters have been optimized, they are stored in a storage device of the computer device 56 in method step 106. Optionally, a further method step 108 (shown here by a dashed line) can follow, in which further measurement data and status information relating to the belt grinder 10 are used to teach the machine learning system 58. The further measurement data can relate to different belt grinders 10, wherein the belt grinders 10 are of a different type than the belt grinder 10.

[0080] at last, Figure 5A method 200 for determining status information relating to a belt grinder 10 is shown. The method 200 is here executed by a computer device 56. In a first method step 202, measurement data relating to the belt grinder 10 are provided to the computer device 56 using the acoustic sensor 44 and the further sensor 50. The provision can be further subdivided into method steps 202a (measurement of the measurement data), 202b (temporary storage of the measurement data (e.g. in a storage device of the belt grinder 10)) and 202c (selective retrieval of the measurement data from the storage device by the computer device 56). In method step 204, status information is determined from the measurement data using the machine learning system 58, in particular the neural network 58a, as a function of the provided or retrieved measurement data. The status information relates to the belt grinder 10 (and / or the grinding belt 18 and / or the workpiece 14 and / or the grinding process). As already mentioned, the method in this exemplary embodiment makes it possible to determine status information that can be selected or predefined by a user of the belt grinder 10. For this purpose, the user can select one of the desired status information by means of an input and output device 60 of the belt grinder 10, which can be connected to the computer device 56 (the method step of selection is implicitly included in the method step 202c - selective call of measurement data). In this case, the user can select between nine taught status information, each of which relates to a different property of the belt grinder 10, the grinding belt 18 used and / or the workpiece 14 and / or the grinding process: a property characterizing the workpiece 14 to be machined; a property characterizing a manufacturing defect on the workpiece 14; a property characterizing the operating mode or operating parameters of the belt grinder 10; a property characterizing an incorrect setting of the belt grinder 10; a property characterizing the load distribution of the belt grinder 10; a property characterizing the degree of wear or wear of the belt grinder 10; a property characterizing the grinding belt 18 used in the belt grinder 10; a property characterizing a blockage and / or dulling of the grinding belt 18; a property characterizing a defect of the grinding belt 18. In method step 206, the result of the evaluation, i.e., the information at least partially based on the determined state information, is then output to a user by means of the input and output device 60. In addition—as shown here in method step 208—the computer device 56 can output control variables to the control device 54 of the belt grinder 10. The control variables are based at least partially on the determined state information.

Claims

1. A method (200) for determining status information relating to a belt grinding machine (10) having at least one grinding belt (18) for grinding a workpiece (14), the method comprising at least the following method steps: providing measurement data relating to the belt grinder (10), wherein at least one acoustic sensor is used to provide the measurement data, determining the state information from the provided measurement data with the aid of a machine learning system (58), wherein the machine learning system (58) is configured to determine the state information based on the provided measurement data, wherein the measurement data provided by the at least one acoustic sensor are provided as input variables to the machine learning system (58), and the machine learning system (58) then determines the state information as an output variable based on a plurality of parameters of the machine learning system (58).

2. The method (200) according to claim 1, wherein the measurement data is also provided using at least one additional sensor (50), wherein the at least one additional sensor (50) is selected from a list of sensors comprising: a sensor for current consumption, an air temperature sensor, an air humidity sensor, a humidity sensor, a distance sensor, a displacement sensor, an imaging sensor, a temperature sensor, a thickness sensor, a torque sensor, a dust load sensor, an inertial sensor, an acceleration sensor, a travel distance sensor, a positioning sensor, a touch-sensitive sensor, and a reflection sensor.

3. The method (200) according to any one of claims 1 to 2, wherein the measurement data is provided selectively.

4. The method (200) of any one of claims 1 to 2, wherein the machine learning system (58) comprises a neural network (58a).

5. The method (200) according to any one of claims 1 to 2, wherein the machine learning system (58) is configured to determine state information relating to at least one of the following properties: Characterizing the properties of the workpiece (14) to be processed; Characterizing properties of manufacturing defects on the workpiece (14); · properties characterizing the operating mode or operating parameters of the belt grinder (10); · a property characterizing an incorrect setting of the belt grinder (10); · properties characterizing the load distribution of the belt grinder (10); Characterizing the degree of wear or the properties of wear of the belt grinder (10); Characterizing the properties of a grinding belt (18) used in the belt grinder (10); Characterizing properties of clogging and / or dulling of the grinding belt (18); • Characterizing properties of defects in the grinding belt (18).

6. The method (200) according to any one of claims 1 to 2, wherein the belt grinder (10) is controlled at least partially based on the determined status information and / or information is output at least partially based on the determined status information by means of an output device (60).

7. The method (200) according to any one of claims 1 to 2, wherein the measurement data is filtered before being provided.

8. The method (200) of claim 1, wherein the at least one sound sensor is an airborne sound sensor and / or a structure-borne sound sensor and / or a vibration sensor.

9. The method (200) of claim 2, wherein the temperature sensor is an infrared sensor or a thermal image sensor.

10. The method (200) according to claim 3, wherein the measurement data are selectively retrieved from the belt grinder (10).

11. The method (200) of claim 7, wherein speech components in the measurement data are filtered out.

12. A method (100) for teaching a machine learning system (58), such that the machine learning system (58) is configured to perform the method according to any one of the preceding claims, comprising the following method steps: providing training data consisting of training input data and training output data, wherein the training input data include measurement data relating to a plurality of status information of the belt grinder (10) and the training output data each include at least one assigned status information relating to the belt grinder (10), wherein the measurement data are provided using at least one acoustic sensor, teaching the machine learning system (58), wherein parameters of the machine learning system (58) are adapted such that the machine learning system (58) determines the respectively associated training output data as a function of the adapted parameters and as a function of the provided training input data, Adding the machine learning system (58) to the computer device (56).

13. The method (100) according to claim 12, wherein in a further method step, further measurement data relating to the belt grinder (10) are provided, wherein at least one predetermined state information relating to the belt grinder (10) is assigned to the further measurement data, and in this way the machine learning system (58) is further taught with the further measurement data.

14. The method (100) according to claim 13, wherein the training input data comprises measurement data and further measurement data relating to a plurality of status information, and the training output data each comprises at least one assigned status information, wherein the measurement data and the further measurement data relate to at least two different types of belt grinders (10), or to at least two belt grinders (10) of the same type with different uses, or to two belt grinders (10) of the same type with the same use.

15. The method (100) according to any one of claims 12-13, characterized in that: The training output data is selected from a list of state information related to at least the following attributes: Characterizing the properties of the workpiece (14) to be processed; Characterizing properties of manufacturing defects on the workpiece (14); · properties characterizing the operating mode or operating parameters of the belt grinder (10); · a property characterizing an incorrect setting of the belt grinder (10); · properties characterizing the load distribution of the belt grinder (10); Characterizing the degree of wear or the properties of wear of the belt grinder (10); Characterizing the properties of a grinding belt (18) used in the belt grinder (10); Characterizing properties of clogging and / or dulling of the grinding belt (18); a property characterizing the defect of the grinding belt (18), or a combination of these attributes.

16. The method (100) according to claim 12, characterized in that: The machine learning system (58) is a neural network (58a).

17. The method (100) according to claim 12, characterized in that: The computer device (56) is a control device (54) of the belt grinder (10).

18. A machine learning system (58) for executing the method according to any one of claims 1-11, which can be obtained by executing the method according to any one of claims 12-17.

19. A computer-readable storage medium having stored thereon a computer program comprising instructions which, when executed on a computer device (56), cause the computer device (56) to perform a method according to any one of claims 1 to 17.

20. A computer device (56) having at least one processor device and a storage device, wherein instructions are stored on the storage device, and when the instructions are executed by the processor device, the instructions cause the computer device (56) to perform the method according to any of the preceding claims 1-17.

21. The computer device (56) according to claim 20, characterized in that The computer device (56) is configured to use the determined state information to determine a control variable for manipulating a physical actuator, and / or to determine a control variable for manipulating an output device (60), and / or to perform a function.

22. The computer device (56) according to claim 21, characterized in that The physical actuator is a belt grinder (10).

23. A belt grinder (10) comprising a grinding belt (18) for grinding a workpiece (14), characterized in that At least one acoustic sensor (44, 44a, 44b, 44c, 44d) for detecting and providing measurement data for carrying out the method according to any one of claims 1-17.

24. The belt grinder (10) according to claim 23, wherein the at least one acoustic sensor (44, 44a, 44b, 44c, 44d) is arranged on or in a grinding pad (12) of the belt grinder (10).

25. The belt grinder (10) according to claim 23 or 24, wherein the at least one acoustic sensor (44, 44a, 44b, 44c, 44d) is arranged in such a manner as to be assigned to a roller (20) of the belt grinder (10).

26. The belt grinder (10) according to any one of claims 23 to 24, wherein the at least one acoustic sensor (44, 44a, 44b, 44c, 44d) is arranged substantially in the center of the grinding pad (12) and / or in the belt grinder (10) with respect to the width (40) of the grinding belt (18) and / or with respect to the width of the grinding pad (12).

27. The belt grinder (10) according to any one of claims 23 to 24, wherein a plurality of acoustic sensors (44, 44a, 44b, 44c, 44d) are arranged at least substantially distributed over the width (40) of the grinding belt (18) and / or over the width of the grinding pad (12) on the grinding pad (12) and / or in the belt grinder (10).

28. The belt grinder (10) according to claim 23, wherein at least two acoustic sensors (44, 44a, 44b, 44c, 44d) are arranged on both sides with respect to the width (40) of the grinding belt (18) and / or with respect to the width of the grinding pad (12) on the grinding pad (12) and / or in the belt grinder (10) and / or are arranged assigned to a roller (20).

29. The belt grinder (10) according to any one of claims 23-24, wherein the at least one acoustic sensor (44, 44a, 44b, 44c, 44d) is connected to a computer device (56) using a gateway (48) and / or to an external computer device (56).

30. The belt grinder (10) according to claim 25, wherein the at least one acoustic sensor (44, 44a, 44b, 44c, 44d) is arranged on a roller suspension of a roller (20) of the belt grinder (10).

31. A grinding pad (12) for use in a belt grinder (10) according to any one of claims 23 to 30, comprising at least one acoustic sensor (44, 44a, 44b, 44c, 44d) for providing measurement data for performing a method according to any one of claims 1 to 17.

32. The method according to any one of claims 1-2 or the belt grinder (10) according to any one of claims 23-24 or the grinding pad (12) according to claim 31, wherein the acoustic sensor (44, 44a, 44b, 44c, 44d) is implemented by a MEMS microphone sensor, or by a piezoelectric sensor, or by a laser microphone sensor.

Images