Field device

Abstract

The invention relates to a field device (10) having a radar sensor (11) for generating radar signals, an antenna (12) for transmitting and receiving the generated radar signals, and a computing arrangement (13) for detecting echo data on the basis of the transmitted and received radar signals, wherein the echo data are indicative of an echo curve (1) of the transmitted and received radar signals, and wherein the echo curve (1) is indicative of a distance measurement value, wherein the computing arrangement (13) is configured to determine, on the basis of the echo data and of reference echo data which are indicative of a reference echo curve (2), a deviation of the echo curve (1) from the reference echo curve (2).

Description

VEGA GRIESHABER KG Our ref.: V11150WO / CGS FIELD EQUIPMENT REFERENCE TO RELATED REGISTRATIONS

[0001] The present application claims priority from German patent application No. 10 2024 125 793.8, filed on September 9, 2024, which is incorporated in full by reference into this document. TECHNICAL AREA

[0002] The invention relates generally to the field of process automation. In particular, the invention relates to a field device, a method for operating a field device, a computer program product, and training data. BACKGROUND

[0003] In industrial measurement technology, particularly in the field of process automation and control, field devices are regularly used to acquire one or more process variables or process measurements, for example, in process vessels or other environments. These process vessels or environments may be used in, or be part of, industrial plant engineering systems, where various process vessels and / or environments are monitored with regard to process variables or process measurements within the framework of process automation and control.

[0004] Field devices include, but are not limited to, flow rate, flow velocity, pressure, differential pressure, temperature, and level measuring devices. Using one or more sensors in a measuring arrangement, field devices typically measure one or more process variables.

[0005] Field devices are often used in industrial plant engineering beyond the applications mentioned above. In particular, field devices with radar sensors, which can be configured to determine distance measurements relative to one or more objects, can be used not only for level and / or limit level measurement, but also in other environments, especially those that are largely or predominantly static with respect to surrounding objects. For example, such field devices can be used in industrial processes, such as water management applications, manufacturing processes, or transportation processes, with robots, cranes, and other moving or mobile units to determine the position of these units within the industrial process.

[0006] When determining a distance measurement, such as a fill level or limit level in a container, particularly using an FMCW radar field device (FMCW: frequency modulated continuous wave radar), a radar signal is typically used. CGS:AFE A signal with a variable frequency is transmitted, reflected by an object, for example, the surface of a substance (in the example of level measurement), and the reflected signal is received and evaluated. A frequency difference between the currently transmitted and the currently received signal can be a measure of the distance, or in other words, the range of an antenna of the field device, in particular the transmitter and receiver of the radar signal, from the object, or in other words, the object, especially its surface. Depending on the distance of the object from the antenna and possibly other factors, signal attenuation can occur, allowing the processing unit of the field device to evaluate a large dynamic range. Furthermore, for example...Internal components in the measurement environment, for example in a container, and static interference signals make the evaluation and determination of a measurement echo (for example level echoes) more difficult, as these can overlap and interfere with the static echoes.

[0007] Such a field device, particularly its radar sensor, can record echo data when receiving the radar signal. This echo data is indicative of the transmitted and received radar signals. After transmission, the radar signal is typically reflected off a surface, such as the contents or other objects in a container, and again captured as an echo signal (the received radar signal). The captured echo signal can thus be compiled into an echo data series, which can, for example, indicate an absolute or relative received signal strength (e.g., in decibels) over a measured distance (e.g., in meters). The echo data series can be graphically represented as an echo curve. The path traveled by the radar signal can also be referred to as a distance measurement or measured distance.The measured quantities can be, in particular, the amplitude of the radar signal, for example, as received signal strength, and the travel time of the outbound and return paths, or the distance, so that the distance between the measuring unit in the form of the antenna, in particular an antenna aperture, and the surface can be determined in a known manner. The amplitude, received signal strength, or signal strength, and the travel time or distance thus yield the echo curve, or, in other words, the signal curve or measurement curve. The echo curve can be converted into a current, the magnitude of which can be, for example, between 4 and 20 mA, with respect to the point(s) of significant amplitude and travel time, and output by the field device as a distance measurement. In a level measurement application, the distance measurement can correspond to the fill level in a container.

[0008] If there is an obstacle on the measuring section, this may be visible in the echo signal or echo curve. The obstacle typically manifests itself as an amplitude corresponding to the obstacle's height in the echo curve.

[0009] Radar sensors typically offer the capability of performing diagnostics via the echo curve. The echo curve allows interference, such as obstacles on the measurement path, to be filtered out, which is visible during commissioning. However, if conditions in the measurement environment change during operation of the field device, the sensor may detect an incorrect signal as the correct one, even if the sensor is initially inactive. When interference signal suppression is applied, the field device may select the wrong echo signal or amplitude as the measurement signal for evaluating the distance measurement, if this signal becomes too large. This can lead to problems, such as process shutdowns, in chemical process plants with tanks where fill levels or limit levels are measured, because the field device delivers an incorrect fill level reading.

[0010] It would be desirable to be able to recognize the possibility of such problems arising at an early stage in order to prevent them in time with appropriate measures. SUMMARY

[0011] The present invention advantageously provides a field device which at least partially eliminates the aforementioned disadvantages.

[0012] The invention is defined in the independent claims. Advantageous further developments and / or embodiments are specified in the dependent claims and in the preceding and following description and disclosure.

[0013] A first aspect of the invention relates to a field device. The field device comprises a radar sensor for generating radar signals. Furthermore, the field device comprises an antenna for transmitting and receiving the generated radar signals. In addition, the field device comprises a processing unit for acquiring echo data based on the transmitted and received radar signals. The echo data is indicative of an echo curve of the transmitted and received radar signals. The echo curve, in turn, is indicative of a distance measurement. The processing unit is configured to determine, based on the (acquired) echo data and reference echo data, which is indicative of a reference echo curve, any deviation of the echo curve from the reference echo curve.

[0014] Advantageously, this refers to a field device whose processing unit is capable of detecting a deviation of a current or measured echo curve from a reference echo curve. The deviation can optionally be further analyzed, particularly by the processing unit, or subjected to analysis, or used directly, for example, output, especially if a minimum deviation exists, as explained in more detail below. In principle, by determining the deviation from the echo curve, it is possible to deduce, in particular, whether an obstacle exists, has developed, or is developing in the measurement path of the transmitted and received radar signals or the antenna, relative to the reference echo curve, for example, because interference echoes are building up in the near-range area of ​​the antenna.Based on this, predictive maintenance can be enabled, as explained in more detail below. For example, the field device can issue a message indicating a change in the echo curve. This allows the field device user to be notified early about maintenance, such as cleaning the antenna if material has accumulated on it that is causing the interference echo. This way, the user can Maintenance should be planned in such a way that it can be carried out, for example, when the process in which the field device is used already has a scheduled downtime. This allows maintenance of the field device to be organized proactively.

[0015] The terms radar sensor, antenna, computing arrangement, radar signals, echo data, echo curve, and reference echo data and reference echo curve are to be interpreted broadly here. Thus, any radar sensor can be used to generate any radar signal of any form. Likewise, any shape or geometry of antenna made of any material can be used, such as, but not limited to, linear antennas and planar antennas. Generally, antennas are understood to be either free-radiating or guided antennas. With a free-radiating antenna or a free-radiating radar-based field device, the radar signals or the emitted waves can be transmitted without contact in the direction of the object to be measured. With a guided antenna or...In a guided radar-based field device, an antenna in the form of a probe, for example a rod or cable, can be used as a waveguide along which the signal can be guided, for example to a medium for level measurement, with the probe being in contact with the medium.

[0016] The computing arrangement can include or be configured as a computing unit, a control unit, and / or an evaluation unit. In particular, the computing arrangement can include or be configured as a computing unit in the form of one or more computers, one or more processor units such as CPUs, one or more microcontrollers, or the like. Specifically, the computing arrangement can be configured to determine, or in other words, calculate, the distance measurement, for example, between the antenna and a medium in a container for level or limit level determination, based on the echo data. The echo data can be evaluated in such a way that a signal strength and / or amplitude of the echo data, especially that of a measurement echo in the echo data or the echo curve, is used to determine the distance measurement.

[0017] Echo data can be in any form or format, for example, as a series of data points, as a function, as the echo curve itself, or similar. The echo curve can be generated graphically based on the echo data, particularly through the computing setup. The echo curve can be understood as mentioned earlier. In particular, the echo curve can generally represent the signal strength and / or amplitude of a radar signal reflected from an object, such as a medium, as a function of the distance to the field device, especially the antenna. The terms "received radar signal" or "echo signal" can refer to the entire signal, including all reflections or echoes. A distinction can be made between reflections by dividing them into measurement echoes and noise echoes, whereby, according to this definition, the measurement echo and the noise echo are based on the same transmitted radar signal.Measurement echoes are the echoes or reflections that indicate the object being measured, while interference echoes may be based on interference or at least do not indicate the object being measured.

[0018] The reference echo data and reference echo curve can essentially be structured in the same way as the echo data and echo curve, meaning they can exist in any form and format. However, the echo data can specifically be the most recently measured or currently measured echo data, or be based on current or recent echo signals. In contrast, the reference echo data serves as a reference, which, as explained in more detail below, can be based on one or more previous measurements and / or be predetermined, for example, predefined for the type of field device and / or one or more applications, such as level measurement and / or specific geometries of objects in the measurement environment. The reference echo data can, for example, be permanently stored on a data storage arrangement of the field device.Echo data can also be stored on the data storage array, at least temporarily, and deleted at predetermined intervals, for example, or transferred to a remote data processing system to keep data storage requirements low.

[0019] The field device can also include a communication arrangement for wired and / or wireless communication with a data processing system, where the data processing system can comprise one or more data processing devices in the form of servers or cloud-based computers. For example, the data processing system can be implemented as a control system for multiple field devices. Wireless communication can be achieved using any transmission technology, such as WLAN, cellular networks, etc. Through wireless communication with the data processing system, the field device can transmit the acquired echo data and / or the measurement data or distance measurements derived from it to the data processing system, where they can then be used, for example, for process automation or process control of a higher-level process.Alternatively, or on a display on the field device itself, the echo data and / or measurement data can be displayed. Furthermore, the field device can have a power supply unit, such as a primary or secondary battery, or at least be equipped to accommodate one, so that the field device can operate autonomously.

[0020] Furthermore, the field device may have a mounting arrangement. The mounting arrangement may include any number of mounting sections and / or fastening means, such as one or more threads, a flange, a rotary fastener, and the like. In particular, the mounting arrangement of the field device may include a mounting section corresponding to a process connection of a process vessel, which may, for example, include a thread (e.g., an internal thread), with a matching thread (e.g., an external thread).

[0021] The detection of a deviation can, for example, take the form of anomaly detection, where the deviation represents an anomaly between the (current and / or measured) echo curve and the reference echo curve and is detected. This detection can be purely qualitative, i.e., whether a deviation exists or not. However, the detection can also It can be quantitative in nature, for example, how large the deviation is. This also allows for an evaluation of the deviation.

[0022] The computing arrangement can be configured to determine the deviation of the echo curve from the reference echo curve in a near-range area of ​​the antenna that is distant from a measuring distance area of ​​the antenna. The near-range area of ​​the antenna and / or radar sensor can, for example, be in the range up to 30%, particularly up to 20%, and especially up to 15% or 10% of the maximum measuring distance or range of the radar signals. The near-range area can be located relatively near the beginning of the measuring path, and the measuring distance area can be located relatively farther away in the direction of propagation of the radar signals or their waves. In principle, however, the measuring distance area can be variable depending on the position of the object being measured, for example, the fill level of the material. Accordingly, the measuring distance area can also be in a long-range and medium-range area, but also in the near-range area.At the time the deviation is detected, the measurement distance range may be far from the near-distance range to allow for a precise determination of the deviation. Deviations can occur particularly frequently in this near-distance range, often due to common causes, such as material adhesion, for example, to the antenna, as mentioned below. Therefore, detection here is especially advantageous for identifying any developing problems with distance measurement at an early stage.

[0023] The computing arrangement can be configured to detect the deviation of the echo curve from the reference echo curve as an echo deviation different from a measured echo, in particular as a distant interference echo, where the measured echo is indicative of the distance measurement value and the echo deviation, in particular the interference echo, is indicative of a source of interference. The echo deviation can manifest itself in various forms, for example as a reduction, a fluctuation, noise, or as the described distant interference echo. This means that the echo deviation can fundamentally consist of a change in the echo curve. The source of interference is explained in more detail below by way of example, for instance, but not limited to, material adhesion or condensation.The desired distance measurement, particularly relative to the surface of an object being measured (e.g., a fill material or filling medium), can be determined from the measurement echo. Conversely, a distant interference echo can disrupt the determination of the distance measurement, especially if the field device or computing system mistakenly interprets the interference echo as a measurement echo. This is particularly true if the interference echo is located near or approaching the measurement echo, for example, because the obstacle causing the interference echo (e.g., material adhering to the antenna) is growing larger and thus encompassing a greater portion of the measurement distance. Advantageously, the computing system can distinguish between the different echoes and their echo signals, thereby determining that the deviation is due to interference. The interference echo can be located particularly in the near-field range. The measurement echo can be located particularly in the measurement distance range.

[0024] For example, the computational arrangement for detecting interference echoes can take into account a spatial shift of the measurement echo in the echo curve over time. In other words, the computational arrangement can consider the change in the echo curve over time and take into account whether and to what extent the measurement echo shifts spatially on the echo curve or measurement curve over time. In particular, the computational arrangement can be configured to consider the spatial shift of measurement echoes and / or interference echoes in the echo curve over time, that is, the echo signals recorded over time. This allows the detection and / or evaluation of deviations to be particularly useful under suddenly changing conditions, or to achieve high accuracy in such cases. For example, if the measurement echo moves into the short-range area, the computational arrangement can detect this based on the last recorded echo signals.The device detects changes in the echo curve over time. This prevents a measured echo from being mistakenly identified as a disturbance echo. However, if an echo suddenly occurs at close range, and the measured echo was previously located at a great distance in the echo curve, the field device can correctly identify the deviation as a disturbance echo.

[0025] The computing system can be configured to generate an indication when a deviation of the echo curve from the reference echo curve is detected, which serves as an indicator for predictive maintenance of the field device. This indication facilitates predictive maintenance by alerting the user to the deviation, allowing them to identify a potential problem early on. The indication can take any form, such as a computer or control signal to generate a notification or warning, a warning or informational message (e.g., with text and / or graphics), or be designed in any other way.

[0026] The computing system can be configured to generate an indication when a predefined minimum deviation of the echo curve from the reference echo curve is detected. Accordingly, a sensitivity for this detection can be set based on the specified minimum deviation. This ensures that only relevant or significant deviations are used to generate the indication and thus for predictive maintenance. Particularly in the case of obstacles that build up over time, such as material buildup, this ensures that an early, but not excessively early, warning can be generated once a certain threshold of the obstacle's size is reached. This allows for the scheduling of maintenance of the field device, such as comprehensive cleaning, in preparation for an impending process shutdown. The minimum deviation can be defined, for example, as a value for an amplitude and / or measuring distance.Distance or extent must be specified.

[0027] The field device can be configured to transmit the indication to a remote data processing system and / or display the indication on a display device of the field device. For this purpose, the field device can, for example, use the communication protocol. Accordingly, the data processing system can centrally output the indication for several monitored field devices or use it to plan predictive maintenance. Alternatively or additionally, the indication can be displayed, for example, as a warning or information message on a display device of the field device. The display device can be of any form, such as a screen, touchscreen, one or more lights (e.g., LEDs), or the like. Accordingly, the indication can be output graphically and / or by the illumination of one or more lights.

[0028] The computing system can be configured to identify a detected deviation as material adhesion and / or condensation on the antenna or on an object reflecting the transmitted radar signals. The causes or obstacles of material adhesion and / or condensation are chosen here only as examples. For instance, the computing system can be designed to recognize or determine known echo characteristics, such as amplitude, for various obstacles or causes, including at least material adhesion and / or condensation, particularly using an algorithm based on machine learning (ML) and / or artificial intelligence (AI), as explained in more detail below.The computing arrangement can also be used to locate the cause or obstacle, particularly relative to the measuring section and / or according to the type of object, for example, a stirrer in a container. Material adhesion is understood here as the adhesion of a material to a surface of the antenna and / or an object, such as a wall in a container, an object within the container (e.g., a stirrer), or the like. The material may be located in a measuring environment, for example, originating from the environment and / or resulting from the process being monitored. For example, the material in a container used for level measurement may originate from chemical processes affecting the material, such as heating, evaporation, stirring, and the like, and thus reach the antenna and / or object.Condensation on the antenna and / or the object can also cause interference by reflecting radar signals and thus generating an echo signal from the field device.

[0029] The field device may have a data storage arrangement on which the reference echo data is stored. The reference echo data may be based on a reference measurement of the field device during its operation, particularly in the field or in application, and / or be predefined reference echo data intended for one or more use cases of the field device. In principle, the reference echo data may also be indicative of several reference echo curves or a bandwidth or range of values ​​of a reference echo curve. Furthermore, the reference echo data may be based on multiple reference measurements. Reference measurements can be taken, in particular, during the initial commissioning or installation of the field device in the respective application or in the field, where it can at least be assumed that no deviations or interference echoes are present. This allows the corresponding reference echo data to be stored and set as a reference for future measured or recorded echo data. Furthermore, the applications may contain certain objects, such as internal components like agitators or similar items in a container, which can generate echo signals. Advantageously, this can be reliably taken into account in the reference echo data, unlike with predefined reference echo curves.

[0030] The computing arrangement can be configured to suppress a noise-laden area of ​​the echo curve and to detect deviations of the echo curve from the reference echo curve outside of this noise-laden area. In particular, a short-range area, especially a limited short-range area or initial short-range area, can be suppressed, which, for example, cannot be used to determine the distance measurement. On the other hand, this measure alone may not guarantee that no noise echoes occur outside of the suppressed, noise-laden area in the echo curve. Advantageously, the computing arrangement can detect deviations outside the suppressed noise-laden area in order to capture them and, furthermore, to utilize the additional measure of suppressing the noise-laden area to prevent incorrect distance measurement determinations.

[0031] The computational setup can be configured to determine the deviation of the echo curve from the reference echo curve by comparing the echo data with the reference echo data. This allows the deviation to be determined relatively easily. Additionally or alternatively, other computational operations besides comparison are possible, whereby "comparison" is to be understood broadly here, encompassing any comparison, comparison, or relating between the echo data and the reference echo data.

[0032] The computing setup can be configured to determine the deviation of the echo curve from the reference echo curve using an algorithm based on machine learning (ML), artificial intelligence (AI), and / or deterministic principles. This can include the aforementioned matching or other operations of the algorithm, particularly in a manner typical of ML and / or AI algorithms, i.e., not deterministic, but rather, for example, through the use of learned structures or determination methods, especially those learned through machine learning, and / or based on neural networks, deep learning, or similar approaches. Alternatively or additionally, a deterministic approach is also possible, as described in more detail below.The algorithm itself can access, incorporate, or use the reference echo data, for example, by being trained using it as training data or taking it into account in its determination. It is therefore not necessary for the reference echo data to be separate from the algorithm; rather, the algorithm data and the reference echo data can exist together, for example, formed by a data set or data structure and stored on the data storage arrangement. The algorithm or algorithms can, in particular, be implemented as a machine learning (ML) and / or computer intelligence (CL) model or computing module, which can, for example, be a computer program. The algorithm can be executed by running the module, especially by the computing arrangement. The algorithm can also be used to determine the distance measurement, or alternatively or additionally, a deterministic determination can be performed. In the variant of deterministic distance measurement with additional determination by the CL and / or ML algorithm, a plausibility check can, for example, be carried out.

[0033] The computing system can be configured to evaluate the deviation using an algorithm based on machine learning (ML), artificial intelligence (AI), and / or determinism. In particular, the algorithm can be the same one used to detect the deviation. Accordingly, the algorithm can be used for both detecting and evaluating the deviation. Especially when evaluating the deviation, the ML- and / or AI-based algorithm offers many advantages, for example, when it comes to qualifying or determining the nature or type of the deviation or the underlying obstacle, such as material adhesion and / or condensation.An algorithm trained using machine learning and / or computer science can advantageously determine the qualification or identification of the type of deviation with high reliability across various applications and environmental conditions. The deviation, once detected, can also be referred to as an anomaly. Differentiation between types of anomalies can be achieved, for example, based on previous measurement data and / or defined by a user. The algorithm can be trained for various applications and environmental conditions, as well as for different measurements and measurement scenarios, and can therefore know or determine the most probable type of deviation for a multitude of situations.A suitably trained algorithm can also reliably determine the quantification and / or localization of deviations on a specific surface or object within the application, such as stirring in a container, along the measurement path, and so on. However, alternative or additional deterministic evaluation and / or manual evaluation by the user or a technician can also be employed. Particularly with a self-learning algorithm and / or one that learns on the field device, it may be possible for the algorithm to receive or consider user input from the user or technician. For example, the user or technician could specify certain interference echoes or deviations for the algorithm to learn from. or without manual or deterministic evaluation regarding their nature, type, location, or the like. This can make the algorithm even better suited for performing evaluations, because, for example, a trained technician can deliberately influence the learning process to obtain an algorithm with high evaluation reliability.

[0034] The algorithm can be configured to compare and process multiple parameters extracted from the echo data and reference echo data in order to determine the deviation.

[0035] The processing of the compared extracted parameters can include a comparison with one or more threshold values ​​assigned to the deviation and / or the different anomaly types. For example, a threshold value for the extracted parameters can be used, such as the difference between the extracted parameters of the echo curve and the reference echo curve. When this difference reaches the threshold, the deviation can then be determined. Different threshold values ​​can also exist for different anomaly types, and / or the reference echo curves can be saved as those exhibiting a specific anomaly type.If, for example, anomalies of a known anomaly type have occurred previously, the measured echo curves can be stored as reference echo curves known to be associated with that specific anomaly type, such as adhesion in the immediate vicinity or condensation in a measuring container. Later, by comparing the extracted parameters from a current echo curve with the reference echo curve(s), it is possible not only to determine whether a deviation or anomaly exists, but also, if applicable, to identify the specific type of anomaly or its cause.

[0036] The algorithm can also be configured to compare and process multiple parameters extracted from the echo data and reference echo data to determine the deviation and identify one of several different anomaly types that is the cause of the echo curve's deviation from the reference echo curve. As described, for example, previous echo data with known anomaly types or causes of deviation can be stored as reference echo data along with their extracted parameters, which are known to be associated with the identified anomaly type.

[0037] Furthermore, the field device can be configured to assign one or more detected deviations to different anomaly types based on user input. In other words, a user can define different anomaly types for one or more detected deviations by entering information on the field device itself or on a connected computer. For example, if a deviation or anomaly is detected that is unknown but exhibits certain extracted parameters in the echo curve, the user can define this as the anomaly type, for example, by manually determining the cause of the deviation, such as adhesion in a nearby area. If the same or similar extracted parameters are then detected, the user can define the anomaly type accordingly. If parameters are detected in subsequently recorded echo curves, they can be output as the specific anomaly type, allowing the user to identify the error and determine the necessary corrective actions. Alternatively, an automated assignment based on previous measurements or determinations can be performed. For example, a database of reference echo data with different parameter sets for various anomaly types can be used to determine the anomaly type of a currently measured echo curve by comparing or matching threshold values.

[0038] The various anomaly types that can be distinguished may include at least two, at least three, at least four, or more (and in particular any combination) of the following: deviation due to reflection, deviation due to a near-field disturbance, deviation due to a flow of material, deviation due to an agitator, deviation due to the formation of a separating layer, deviation due to condensation, deviation due to adhesion, deviation due to a change in a process through which the radar signals pass, and deviation due to a change in a material through which the radar signals pass. Accordingly, a wide variety of anomaly types could be distinguished in order to provide the field device user with information about the nature of the deviation or anomaly when the corresponding anomaly type is displayed.

[0039] The field device can also be configured to ignore one or more anomaly types based on user input. Accordingly, in addition to or as an alternative to defining various anomaly types for detected deviations, the user can also define anomaly types that are not of interest to them, for example, because they are unavoidable or undesirable in the process, or because they are faulty.

[0040] The extracted parameters can include at least one, at least two, or more (in particular, any combination) of the following: the slope of one or more peaks in at least one sub-region of the echo curve and reference echo curve, the height of one or more peaks in at least one sub-region of the echo curve and reference echo curve, the number of peaks in at least one sub-region of the echo curve and reference echo curve, the distance between peaks in at least one sub-region of the echo curve and reference echo curve, and the noise level in at least one sub-region of the echo curve and reference echo curve. "In at least one sub-region" can mean, for example, a near-range, a mid-range, and / or a far-range, or overall, the entire distance range.The extracted parameters can be specified accordingly for the different areas in order to achieve an even better resolution for the identification of different anomaly types.

[0041] In particular, the ML and / or AI-based algorithm can be an unsupervised learning algorithm. Such algorithms operate without explicit labels and discover patterns or anomalies in the data themselves and can also be used as Unsupervised Representation Learning is a broader term. For example, algorithms such as OneClassSVM and autoencoders can be used.

[0042]

[0043] In particular, the deterministic algorithm can be a distance-based algorithm. For example, anomaly detection can be performed deterministically using distance calculations. The underlying principle is that normal data points or parameters are usually located close to each other. Anomalies, on the other hand, are unusual data points or parameters that are far removed from the others. The Euclidean distance, for instance, can be used as a measure of distance.

[0044]

[0045] The algorithm can be configured to evaluate the deviation in order to identify a detected deviation as material adhesion and / or condensation on the antenna and / or on an object from which the transmitted radar signals are reflected, to identify at least one object from which the transmitted radar signals are reflected that causes the detected deviation, to determine error data associated with the detected deviation that are indicative of an error risk associated with the detected deviation, and / or Maintenance data for certain which are indicative of a maintenance measure and / or a maintenance time to rectify the deviation by servicing the field device (10). This allows the algorithm to identify the type of deviation, as previously mentioned. Furthermore, it can additionally or alternatively identify the type of object, such as a container wall, a stirrer, or any other type within a specific measurement application or environment. Additionally or alternatively, it can determine error data that indicates the risk of failure. This risk can be associated with the probability of a future problem with the field device and may indicate when this problem might occur. The problem could, for example, be a failure of the field device's measurement operation due to the deviation, or it could be inaccurate distance measurements. Users can utilize this information to proactively plan maintenance, particularly during downtime of the application or process.

[0046] The algorithm, based on machine learning and / or artificial intelligence, can be a pre-trained algorithm stored on a data storage system. Alternatively, the algorithm can be executed on a remote data processing system. Since the algorithm is pre-trained, or in other words, learned, and therefore immediately usable, execution can be performed directly by the computing system without requiring it to be executed on a geographically remote system. to relocate the data processing system. Because the algorithm is pre-trained, a very comprehensively trained algorithm can be provided. This means that the algorithm's training data covers many different use cases, measurement environments, deviations, especially at different locations and / or of varying magnitudes, and the like, and the algorithm can therefore be very well trained to reliably detect and / or evaluate deviations.

[0047] The computing system can be configured to train the machine learning and / or artificial intelligence-based algorithm during the operation of the field device. The algorithm can be pre-trained or untrained when it is first used or trained. A pre-trained algorithm can be further trained more effectively, particularly in the application and under the operating conditions to which the field device will ultimately be exposed in the field. An untrained algorithm can be trained in this way. Typically, the computing power required for training during operation, or in other words, the runtime of the field device, is greater than the computing power required to execute the algorithm to detect and / or evaluate detected deviations.Therefore, in this variant, it can be advantageous to use a computing system with comparatively greater processing power than in the variant without training the field device during runtime. It can also be explicitly stipulated that the algorithm is not trained during the operation of the field device, or that the computing system is not configured to train the algorithm during operation, in order to provide a comparatively more cost-effective computing system.

[0048] The radar sensor, for example, does not necessarily have to be a continuous-wave radar sensor. In other words, a frequency-modulated continuous-wave radar (FMCW) field device can be used. This allows a radar signal with a variable frequency to be transmitted, received, and analyzed.

[0049] The field device can be configured, in particular but not limited to, a level and / or limit level measuring device for measuring the level and / or limit level of a medium in a container based on echo data. For this purpose, the field device can be mounted in or on the container, particularly by means of the aforementioned mounting arrangement. Distance measurements in other applications, such as industrial processes like manufacturing or transportation, are also possible. These applications can utilize, for example, robots, cranes, and other movable or mobile units to determine the position of the units within the industrial process.

[0050] A second aspect of the invention relates to methods for operating a field device, wherein the method comprises: Sending and receiving radar signals generated by a radar sensor of the field device using an antenna of the field device, - Acquisition of echo data based on the transmitted and received radar signals, wherein the echo data are indicative of an echo curve of the transmitted and received radar signals, and wherein the echo curve is indicative of a distance measurement, and - Determining a deviation of the echo curve from the reference echo curve based on the echo data and reference echo data that are indicative of a reference echo curve.

[0051] The method may, in particular, explicitly include the generation of the radar signal by means of the antenna. Furthermore, the method may include any of the features or steps mentioned herein that are described in relation to the first aspect of the invention or the field device, or that may apply analogously thereto.

[0052] The procedure can, for example, include evaluating the detected deviation. For instance, the deviation of the echo curve from the reference echo curve can be determined in the antenna's near-field range. Furthermore, the deviation of the echo curve from the reference echo curve can be identified as a disturbance echo located away from the measured echo. The detection of the disturbance echo can also take into account a spatial shift of the measured echo in the echo curve over time. In addition, an indication can be generated, particularly upon detection of a predefined minimum deviation. This indication can then be transmitted to the remote data processing system, especially wirelessly and / or via a wired connection, and / or displayed on the field device's display. The detected deviation can also be identified as material adhesion and / or condensation.It is also possible to suppress a noisy area of ​​the echo curve. The echo curve can be compared to the reference echo curve.

[0053] The procedure may further include: using and / or training an algorithm based on machine learning and / or artificial intelligence to detect and / or evaluate the deviation. The algorithm may also be pre-trained. The deviation may occur, for example, as described herein.

[0054] A third aspect of the present invention relates to a computer program product comprising instructions that, when the program is executed by a computer, cause it to execute the method according to the second aspect of the present invention. The computer can, for example, be the computing arrangement of the field device.

[0055] A fourth aspect of the present invention relates to training data for training a machine learning and / or artificial intelligence-based algorithm for use by a field device, comprising a radar sensor for generating radar signals, an antenna for transmitting and receiving the generated radar signals, and a computing arrangement for acquiring echo data based on the transmitted and received radar signals, wherein the echo data are indicative of an echo curve of the transmitted and received radar signals, and wherein the echo curve is indicative of a distance measurement, wherein the computing arrangement is configured to determine, based on the echo data and reference echo data indicative of a reference echo curve, a deviation of the echo curve from the reference echo curve. To determine the echo curve, with the training data based on the echo data and the reference echo data to train the machine learning and / or artificial intelligence-based algorithm to detect and / or evaluate the deviation.

[0056] The computer program product and the training data may include any of the features or steps mentioned herein that are described in relation to the first aspect of the invention or the field device or the second aspect of the invention or the method, or that may apply analogously thereto.

[0057] For example, the training data can be at least partially generated during the operation of the field device. Alternatively or additionally, the training data can be at least partially generated outside the field device in which it is used, for example, training data generated in a large number of field devices in the same or different applications. The training data can also include input data that identifies and / or characterizes the deviation, with the input data being based, for example, on a manual evaluation of the cause or obstacle of the deviation during the maintenance of a field device. The input data can thus characterize the identified deviations with respect to the actual obstacles that cause them, for example, by the nature, type, size, and other characteristics of the obstacles. For example, the input data can also or alternatively include instructions specifying certain deviations and / or evaluations that the algorithm should use for learning. The input data could, for example, come from a technician. BRIEF DESCRIPTION OF THE FIGURES

[0058] Exemplary embodiments of the invention are described below with reference to the accompanying figures.

[0059] Fig. 1 shows a schematic view of a field device during an application.

[0060] Fig. 2 shows another schematic view of the field device from Fig. 1.

[0061] Fig. 3 shows a schematic view of an exemplary reference echo curve of the field device of Figs. 1 and 2.

[0062] Fig. 4 shows a schematic view of an exemplary echo curve of the field device of Figs. 1 and 2.

[0063] Fig. 5 shows a schematic view of another exemplary echo curve of the field device of Figs. 1 and 2.

[0064] Figures 6a and 6b show schematic views of methods for operating the field device from Figures 1 and 2 using different algorithms.

[0065] Similar, similar-looking, identical or equivalent elements in the figures may be provided with similar or identical reference symbols. DETAILED DESCRIPTION OF EXECUTION FORMS

[0066] Figure 1 schematically shows a field device 10, which is designed as a radar level gauge. The field device 10 in Figure 1 is shown by way of example in an application for measuring the fill level and / or limit level of a substance 100, for example a filling medium, in a container 20. For this purpose, the field device 10 can be attached to and / or inside the container 20. Other applications are possible.

[0067] An antenna 12 inside the container 20 is shown here, by way of example, directed towards a surface of the contents 100 to transmit and receive a radar signal, as indicated by two arrows in Fig. 1. The radar signal can be generated by a radar sensor 11 of the field device 10 (see Fig. 2) and transmitted via the antenna 12 and received back after reflection from the contents 100, in particular as an echo signal from which an echo curve 1 can be generated, as shown by way of example in Figs. 4 and 5.

[0068] As shown schematically in Figure 2, the field device 10 can include further arrangements or components, for example, a computing arrangement 13, which can be configured to evaluate the transmitted and received radar signals or the echo signal in order to determine a distance measurement between the antenna 12 and the surface of the material 100. In the present application example, the distance measurement represents the fill level of the container 20 containing the material 100 or can be derived from it. The computing arrangement 13 can receive echo data from the radar sensor 11, which is indicative of the echo curve 1 of the transmitted and received radar signals, the echo curve 1 in turn being indicative of the distance measurement.

[0069] An exemplary, measured echo curve 1 is shown in Fig. 4, plotted against a signal strength in dB and a measuring distance in meters. This echo curve 1 has a measurement echo 3 in the form of a peak in a measurement distance range remote from the antenna 12, which is far from a near-distance range of the echo curve 1 close to a range of the antenna 12. The measurement echo 3 is indicative of the distance measurement and can be used to determine the distance measurement by the computing arrangement 13.

[0070] In addition to determining the distance or fill level value, the computing arrangement 13 is configured to detect a deviation of the echo curve 1 from the reference echo curve 2 based on the acquired echo data and reference echo data that are indicative of a reference echo curve 2 (see Fig. 3). The reference echo data can be recorded, for example, initially during commissioning or during early operation of the field device 10 and stored in a data storage arrangement 14 with one or more data storage devices of the field device 10 (see Fig. 2).

[0071] An exemplary reference echo curve 2 is shown in Fig. 3, which is plotted as an echo signal against a signal strength in dB and a measuring distance in meters. The reference echo curve 2 also has a measurement echo 3. Unlike the currently measured or generated echo curve 1, the reference echo curve 2 does not exhibit any interference echoes 4, as is shown by way of example in Fig. 4 for echo curve 1. This may be because, for example, at the beginning of the Typically, when the field device 10 is put into operation, no interference effects or causes of interference, such as material adhesion to the antenna 12 (e.g., dirt), are present. Therefore, the reference echo curve 2 can be used by the computing arrangement 13 as a reference for the echo curve 1 in order to detect deviations in the form of interference echoes 4.

[0072] If, due to intervening disturbances, the echo curve 1 exhibits a noise echo 4 during the operation of the field device 10, as exemplified in Fig. 4, this can be detected by comparing the echo curve 1 with the reference echo curve 2. In addition to or as part of a comparison, it is also possible to use an algorithm based on machine learning and / or artificial intelligence, which can be used to detect the deviation or the noise echo 4 and optionally to evaluate the deviation or the noise echo 4. The algorithm can be executed before the computing unit 13 and stored on the data storage unit 14. The algorithm can optionally be trained during the operation of the field device 10 and / or be pre-trained.

[0073] A display 15 of the field device 10 (see Fig. 2) can be used to output an indication regarding the detection of the deviation or the disturbance echo 4, particularly if a minimum deviation or disturbance echo is present. This can be used as an indication for predictive maintenance. Alternatively or additionally, it is possible to use a wireless or wired communication arrangement 16 of the field device 10 (see Fig. 2) that can transmit the indication or deviation to a data processing system (not shown) located remotely from the field device 10, either to display it there or to notify a user so that predictive maintenance can be planned.

[0074] Figure 5 shows an example of the change over time of an example echo curve 1, in which the interference echo 4 increases in size. This can be monitored by the field device 10 throughout its operating time. By determining the change, in this case the increase, of the interference echo 4, predictive maintenance can be planned even more precisely. For example, the growth rate of the interference echo 4 can be determined by the computer 13 to predict when maintenance might be advisable or necessary, so that the computer 13 can indicate or provide a well-founded recommendation for predictive maintenance.

[0075] Figures 3 to 5 further show the suppression of a (typically) interference-laden area 5 in the near-distance range of the echo curves 1, 2. As can be seen from Figures 4 and 5, this suppression can indeed be carried out by the computing arrangement 13 to exclude any influence of this area 5 on the determined measurement distance value. However, it may be that the suppression alone is not sufficient to exclude possible problems with regard to the interference echoes 4, in particular that the interference echo 4 is used by the field device 10 to determine the distance measurement value instead of the measurement echo 3. Accordingly, it is also advantageous to detect or determine the interference echoes 4 as described in order to prevent to enable predictive maintenance in order to eliminate, for example by cleaning away contaminants as the cause of the interference echoes 4.

[0076] Figure 6a schematically shows the steps of a procedure for operating the field device 10.

[0077] In a first step S1, the method involves the acquisition of echo data indicative of one or more echo curves 1, which are based on the transmission and reception of radar signals generated by the radar sensor 11 of the field device 10 using an antenna 12 of the field device 10. In step S1, reference echo data are also provided, which are indicative of one or more reference echo curves 2.

[0078] In step S2, various parameters are extracted from the echo data and reference echo data, or rather, their respective curves. These are then compared in step S3, using an algorithm based on machine learning and / or computer science, such as OneClassSVM, as shown in the example in Fig. 6a. This involves calculating a score (SampleScore) or value, or alternatively a Euclidean distance, between the parameters of the reference echo data and the echo data. In step S4 of the procedure, this score is compared to a defined threshold. This threshold can also be set in step S4, for example, by a user. In step S5, the extracted parameters and the comparison with the reference echo curves determine whether there is a deviation or no deviation. Accordingly, in step S5, the echo curves can be classified into those with deviations or deviations.exhibiting anomalies, and those which are anomaly-free or unremarkable.

[0079] Figure 6b shows an essentially identical procedure to that shown in Figure 6a, with the difference that in step S3 a distance calculation is performed based on a deterministic algorithm; for example, the Euclidean distance can be calculated. Accordingly, the threshold value in step S4 can also be different.

[0080] In principle, the threshold value can be adjusted manually or automatically during the runtime of the algorithm shown in Figures 6a and 6b. For example, an adjustment can be made if deviations are detected for too many echo curves, especially if a predefined limit of echo curves with deviations is exceeded.

[0081] Accordingly, the presented field device and method provide a way to detect anomalies in a process monitored by the field device using radar signals, such as a process in a container, in order to react to them. A reaction can fundamentally consist of feedback from the field device to the user and / or an adjustment of the field device's signal processing. As described, parameters or features of the echo curve can be extracted and compared using machine learning (such as OneClassSVM, autoencoder, etc.) or distance calculations. A threshold value can be used to differentiate between anomalies and normal echo curves. It is also possible to categorize the results of such an algorithm into two or three groups, indicating whether an anomaly is present, not present, or possibly present, or alternatively, to convert them into a regression-based scale. Anomalies can include, for example, reflections, disturbances in the immediate vicinity of the field device, or... Radar sensors, bulk flow, agitator, separation layer in the medium, condensate, adhesions, cleaning, process and / or material changes, or other deviations from the previous state.

[0082] Various parameters describing the curve can be extracted from one or more reference echo curves. These parameters are also calculated for new curves generated during the current process of the vessel in which the field device is used, such as a stirring process or other chemical process. The distance values, or the so-called sample score, are then calculated using OneClassSVM between the reference echo curves and the new or currently measured echo curves. Accuracy can be increased by including the temporal evolution of the parameters. Based on a threshold value, the curves can be divided into two or more groups, and the process can be adjusted accordingly.

[0083] In addition to a general determination of the presence of an anomaly, anomalies can also be categorized into classes of frequent disturbances based on known parameter patterns and temporal trends (e.g., the last detected anomaly was likely adhesion or condensation). This allows for a more appropriate response to the anomaly.

[0084] An example of a tank cleaning application would be if the radar sensor in the application is measuring but is dirty, for example, due to build-up. The algorithm can trigger an alarm in this case, resulting in the tank being cleaned without removing the sensor. By running the algorithm again, it can be determined whether the cleaning was successful.

[0085] Another example of a potential application is a process involving an agitator, specifically the use of an agitator in a tank. In this scenario, the radar sensor is integrated into the tank along with the agitator. The agitator can be stopped so that its impeller is within the radar's detection range or cone. The algorithm detects an anomaly, and the signal processing can be switched off once the agitator moves outside the radar's field of view.

[0086] Another example of a further application is flow of material. In this scenario, the radar sensor operates, for example, on a gypsum silo. As the silo is filled, a mist / flow of material can occur, which can lead to an inaccurate distance reading from the radar sensor. The signal processing of the field device used to determine the distance can be switched to a more robust mode, and the filling process stops when the echo curves are no longer detected as anomalies, allowing the signal processing to be switched back to its normal state.

[0087] Using only a small amount of reference data, both general and specific operational anomalies can be detected and addressed. Specifically, this is achieved by extracting quickly computable parameters from the echo curves, comparing these parameters to the parameters of the reference echo curves, and differentiating between them using one or more threshold values. The algorithm can be dynamically configured. For example, it can be implemented that the user can declare specific anomalies. In the future, such an anomaly will then be flagged as... These are labeled. Furthermore, it can be provided that a user declares an anomaly as a desired state. Then the anomaly will no longer be reported in the future.

[0088]

[0089] It should be further noted that "comprehensive" and "having" do not exclude other features, and the indefinite articles "a" or "a" do not exclude a plurality. It should also be noted that features described with reference to one of the above embodiments may also be used in combination with other features or steps of other embodiments described above. Reference numerals in the claims are not to be considered limitations.

[0090] As used herein, the expression "indicative of" can mean, for example, "reflective" and / or "comprehensive." Accordingly, a unit, element, and / or step referred to here as "indicative of [...]" may be used synonymously or interchangeably with one, two, or all of the aforementioned units, elements, and / or steps "comprehensive [...]" and the aforementioned unit, element, and / or step "reflective [...]".

[0091] Furthermore, expressions such as "based on," "related," "associated," and similar expressions are not to be understood exclusively in relation to the units, elements, and / or steps to which they refer, unless otherwise specified. Instead, unless otherwise specified, these expressions are to be understood as meaning that, for example, a unit, element, or step to which one of these expressions or a similar expression refers, e.g., "based on" one or another unit, element, or step, does not preclude the possibility that the unit, element, or step in question may also be "based" on a different unit, element, or step than the one to which it refers.

[0092] Any designation of elements as first, second, etc., as indicated herein, serves solely to make the elements referable and distinguishable from one another. The designation of elements in no way constitutes a limitation of the scope of this disclosure.

Claims

REQUIREMENTS 1. Field device (10), comprising: - a radar sensor (11) for generating radar signals, - an antenna (12) for transmitting and receiving the generated radar signals, and - a computing arrangement (13) for acquiring echo data based on the transmitted and received radar signals, wherein the echo data are indicative of an echo curve (1) of the transmitted and received radar signals, and wherein the echo curve (1) is indicative of a distance measurement value, wherein the computing arrangement (13) is configured to determine a deviation of the echo curve (1) from the reference echo curve (2) on the basis of the echo data and reference echo data, which are indicative of a reference echo curve (2).

2. Field device (10) according to claim 1, wherein the computing arrangement (13) is configured to determine the deviation of the echo curve (1) from the reference echo curve (2) in a near-distance range of the antenna (12) that is remote from a measurement distance range of the antenna (12).

3. Field device (10) according to claim 1 or 2, wherein the computing arrangement (13) is configured to determine the deviation of the echo curve (1) from the reference echo curve (2) as an echo deviation different from a measurement echo (3), for example as a distant interference echo (4), wherein the measurement echo (3) is indicative of the distance measurement value and the echo deviation, for example the interference echo (4), is indicative of a disturbance cause.

4. Field device (10) according to claim 3, wherein the computing arrangement (13) for determining the echo deviation, for example the interference echo (4), takes into account a local shift of the measurement echo (3) in the echo curve (1) over time.

5. Field device (10) according to one of the preceding claims, wherein the computing arrangement (13) is configured to generate an indication upon detection of the deviation of the echo curve (1) from the reference echo curve (2), which is indicative for predictive maintenance of the field device (10).

6. Field device (10) according to claim 5, wherein the computing arrangement (13) is configured to generate the indication upon detection of a predetermined minimum deviation of the echo curve (1) from the reference echo curve (2).

7. Field device (10) according to claim 5 or 6, wherein the field device (10) is configured to transmit the indication to a remote data processing system and / or to display the indication on a display means (15) of the field device (10).

8. Field device (10) according to one of the preceding claims, wherein the computing arrangement (13) is configured to identify a detected deviation as material adhesion and / or condensation on the antenna (12) or on an object from which the transmitted radar signals are reflected.

9. Field device (10) according to one of the preceding claims, wherein the field device (10) has a data storage arrangement (14) on which the reference echo data are stored, wherein the reference echo data are based on a reference measurement of the field device (10) during its operation and / or are predetermined reference echo data that are predefined for one or more use cases of the field device (10).

10. Field device (10) according to one of the preceding claims, wherein the computing arrangement (13) is configured to suppress a noise-infected area (5) of the echo curve (1) and to determine the deviation of the echo curve (1) from the reference echo curve (2) outside the noise-infected area of ​​the echo curve (1).

11. Field device (10) according to one of the preceding claims, wherein the computing arrangement (13) is configured to determine the deviation of the echo curve (1) from the reference echo curve (2) by comparing the echo data with the reference echo data.

12. Field device (10) according to one of the preceding claims, wherein the computing arrangement (13) is configured to evaluate the deviation using an artificial intelligence (AI) based on machine learning (ML) and / or a deterministic algorithm.

13. Field device (10) according to one of the preceding claims, wherein the computing arrangement (13) is configured to determine the deviation of the echo curve (1) from the reference echo curve (2) using an artificial intelligence (AI) based on machine learning (ML) and / or a deterministic algorithm.

14. Field device (10) according to claim 13, wherein the algorithm is configured to compare and process several parameters extracted from the echo data and reference echo data in order to determine the deviation.

15. Field device (10) according to claim 14, wherein the algorithm is further configured to compare and process several parameters extracted from the echo data and reference echo data in order to determine one of several different anomaly types that is the cause of the deviation of the echo curve (1) from the reference echo curve (2).

16. Field device (10) according to claim 14 or 15, wherein the processing of the extracted parameters compared with each other comprises a comparison with one or more threshold values ​​that are associated with the deviation and / or the different anomaly types.

17. Field device (10) according to any one of claims 12 to 16, wherein the field device (10) is configured to assign one or more detected deviations to different anomaly types.

18. Field device (10) according to claim 17, wherein the field device (10) is configured to assign the one or more detected deviations to the different anomaly types based on user input.

19. Field device (10) according to any one of claims 15 to 18, wherein the different types of anomalies that can be distinguished include at least two or more of the following: deviation due to reflection, deviation due to near-field disturbance, deviation due to a bulk flow, deviation due to an agitator, deviation due to separation layer formation, deviation due to condensation formation, deviation due to adhesion, deviation due to a change in a process through which the radar signals pass, and deviation due to a change in a material through which the radar signals pass.

20. Field device (10) according to any one of claims 15 to 19, wherein the field device (10) is configured to ignore one or more anomaly types based on user input.

21. Field device (10) according to one of claims 13 to 20, wherein the extracted parameters comprise at least one or more of the following: a slope of one or more peaks in at least a sub-region of the echo curve (1) and reference echo curve (2), a height of one or more peaks in at least a sub-region of the echo curve (1) and reference echo curve (2), a number of peaks in at least one sub-region of the echo curve (1) and reference echo curve (2), a distance between peaks in at least one sub-region of the echo curve (1) and reference echo curve (2), and a noise level in at least one sub-region of the echo curve (1) and reference echo curve (2) 22. Field device (10) according to one of claims 12 to 21, wherein the algorithm based on ML and / or Kl is an unsupervised algorithm.

23. Field device (10) according to one of claims 12 to 22, wherein the deterministically based algorithm is a distance-based algorithm.

24. Field device (10) according to one of claims 12 to 23, wherein the algorithm is configured to evaluate the deviation in order to identify a detected deviation as material adhesion and / or condensation on an object from which the transmitted radar signals are reflected.

25. Field device (10) according to one of claims 12 to 24, wherein the algorithm is configured to evaluate the deviation in order to identify at least one object from which the transmitted radar signals are reflected which cause the detected deviation.

26. Field device (10) according to one of claims 12 to 25, wherein the algorithm is configured to evaluate the deviation in order to determine error data associated with the detected deviation which are indicative of an error risk associated with the detected deviation.

27. Field device (10) according to one of claims 12 to 26, wherein the algorithm is configured to evaluate the deviation in order to determine maintenance data which are indicative of a maintenance measure and / or a maintenance time to correct the deviation by maintaining the field device (10).

28. Field device (10) according to any one of claims 12 to 27, wherein the machine learning and / or artificial intelligence-based algorithm is a pre-trained algorithm stored on a data storage arrangement (14) of the field device (10).

29. Field device (10) according to any one of claims 12 to 28, wherein the computing arrangement (13) is configured to train the ML and / or Kl-based algorithm during the operation of the field device (10).

30. Field device (10) according to one of the preceding claims, wherein the radar sensor (11) is designed as a continuous wave radar sensor.

31. Field device (10) according to one of the preceding claims, wherein the field device (10) is designed as a level and / or limit level measuring device for measuring a level and / or limit level of a medium in a container (20) based on the echo data (1).

32. Method for operating a field device (10), wherein the method comprises: - Sending and receiving radar signals generated by a radar sensor (11) of the field device (10) using an antenna (12) of the field device (10), - Acquisition of echo data based on the transmitted and received radar signals, wherein the echo data are indicative of an echo curve (1) of the transmitted and received radar signals, and wherein the echo curve (1) is indicative of a distance measurement value, and - Determining a deviation of the echo curve (1) from the reference echo curve (2) based on the echo data and reference echo data that are indicative of a reference echo curve (2).

33. The method of claim 32, wherein the method further comprises: - Using and / or training a machine learning and / or artificial intelligence-based algorithm to detect and / or evaluate the deviation.

34. Computer program product comprising instructions which, when the program is executed by a computer, cause it to execute the method according to claim 32 or 33.

35. Training data for training a machine learning (ML) and / or artificial intelligence (AI) based algorithm for use by a field device (10), comprising a radar sensor (11) for generating radar signals, an antenna (12) for transmitting and receiving the generated radar signals, and a computing arrangement (13) for acquiring echo data based on the transmitted and received radar signals, wherein the echo data are indicative of an echo curve (1) of the transmitted and received radar signals, and wherein the echo curve (1) is indicative of a distance measurement, wherein the computing arrangement (13) is configured to detect, based on the echo data and reference echo data indicative of a reference echo curve (2), a deviation of the echo curve (1) from the reference echo curve (2).where the training data is based on the echo data and the reference echo data for training the ML and / or KL-based algorithm to detect and / or evaluate the deviation.

Images