Field device for detecting process parameters
By combining the training mechanism of deterministic computing module and AI computing module in field devices, the problem of insufficient flexibility of the equipment for specific uses and environmental changes is solved, and high-precision and reliable process parameter detection is achieved.
Patent Information
- Application Number
- CN202410670088.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2023-06-05
- Filing Date
- 2024-05-28
- Publication Date
- 2024-12-06
AI Technical Summary
Artificial intelligence modules in existing field devices are not flexible enough for specific uses and environmental changes, and it is difficult to provide reliable process parameter measurement results.
Design a field device equipped with sensor devices and computing devices, which are used to detect measurement signals of process parameters. The computing device includes a deterministic computing module and an AI-based AI computing module, allowing training of the AI computing module during the operation of the device. Parameterized sensor devices according to specific use and environment to ensure the robustness of the AI module and high measurement accuracy.
It realizes the reliability and flexibility of the AI computing module, can adapt to specific applications and environmental changes, provide more accurate process parameter measurement results, and save resources when necessary, and improve detection accuracy.
Smart Images

Figure CN119085738A_ABST
Abstract
Description
Technical Field
[0001] The present invention generally relates to the field of process automation. In particular, the present invention relates to a field device for detecting process parameters, such as a level measuring device for detecting the level of a medium, in particular a limit level measuring device. The present invention also relates to a method for training an AI computing module based on artificial intelligence (AI) of a field device, a method for operating a field device, a computer program product associated with the method, and training data for training an AI computing module. Background Art
[0002] In industrial measurement technology, especially in the field of process automation and process control, field devices are often used to detect one or more process variables or process parameters. In particular, field devices include flow measuring devices, flow velocity measuring devices, pressure measuring devices, pressure difference measuring devices, temperature measuring devices and level measuring devices. By means of corresponding sensors or corresponding sensor devices, field devices usually detect measurement signals related to one or more process parameters. Then, based on the measurement signals and / or based on the evaluation of the measurement signals, the measured values of the corresponding process parameters can be determined via a computing unit, a control unit and / or an evaluation unit of the corresponding field device.
[0003] In order to perform arithmetic operations during the operation of field devices (e.g., arithmetic operations during the actual determination of measured values or arithmetic operations during diagnostic processes), field devices have been proposed which use artificial intelligence to perform these arithmetic operations. The artificial intelligence is pre-trained or trained on the basis of a large number of field devices. Such artificial intelligence in field devices is inflexible or barely flexible for the specific use of the field device. Summary of the invention
[0004] In the context of the present invention, an improved field device for detecting a process parameter can be provided in an advantageous manner, in particular for respective specific uses of the field device.
[0005] The invention is defined in the independent claims. Advantageous developments and / or embodiments are specified in the dependent patent claims and in the following description.
[0006] A first aspect of the invention relates to a field device for detecting a process parameter. The field device can be designed in particular as a level measuring device for detecting the level of a medium. The field device has a sensor device, which is configured to detect a measurement signal related to the process parameter. The field device also has a computing device, which is configured to determine a measured value of the process parameter based on the measurement signal. The computing device has a deterministic computing module and an AI computing module based on artificial intelligence. The field device is configured to train the AI computing module during its operation.
[0007] Thus, a field device is advantageously provided which allows the training of the AI computing module during the operation time of the field device or in other words during the run time. Operation refers to the measuring operation of the field device detecting process parameters. Thus, in particular, the AI computing module is trained based on a specific use or based on parameterization during operation of the field device in a specific use. Thus, the AI computing module is robust to parameter changes that may be caused by aging or application changes. Unlike pre-trained field devices, it can be ensured that the AI computing module of the field device provides reliable results for a specific application.
[0008] In this context, a specific use is here particularly understood as the corresponding use of the field device for a specific application and / or environment. For example, the field device can be used for different applications and exposed to different environments. In order to determine the accurate measurement value, the field device, in particular the sensor device, can be parameterized for different applications and / or environments to allow high measurement accuracy. Parameterization can be particularly understood as characterizing the performance of the field device by affecting the parameters of the process parameter detection. Parameterization can at least include the parameters of the field device itself, i.e., the specific settings of the field device, such as measurement sensitivity. Alternatively or additionally, parameterization can also include external parameters such as measurement conditions (e.g., the density of the medium to be measured, the volume of the container containing the medium, etc.) and alternative environmental parameters such as ambient temperature, ambient pressure, etc.
[0009] The terms "computing means" and "computing module" should be interpreted broadly. In particular, a computing means may refer to a computer environment comprising at least one computing unit for performing arithmetic operations. Such a computer environment or such a computing unit may include a processor for performing arithmetic operations. The computing means or the computer environment may optionally include other components, such as one or more data memories. On the other hand, a computing module may represent a computer program product itself or a combination of a computer program product and a computing unit for executing the computer program product. By executing a computing module, a computing operation comprising one or more arithmetic operations may be performed in particular.
[0010] The deterministic calculation module may be configured to perform at least one deterministic arithmetic operation. The at least one deterministic arithmetic operation of the deterministic calculation module may be used to determine at least a measured value of a process parameter based on the measurement signal. The term "deterministic" may refer here and hereinafter to the first calculation module having a deterministic algorithm and / or being designed to perform the deterministic algorithm, in which well-defined and / or reproducible states occur and / or a specific input always leads to the same output when passing through the same state.
[0011] On the other hand, the AI computing module can be configured to perform at least one AI-based arithmetic operation, in particular at least one non-deterministic, probability-based and / or classification-based arithmetic operation. Non-deterministic, probability-based and / or classification-based arithmetic operations may refer to arithmetic operations whose results may be outputs of one or more probabilities or probability values of one or more classes of events and / or results, for example. Alternatively or additionally, the output may be one or more classes and / or class names. Therefore, the AI computing module may have a classifier and / or be designed as a classifier. In addition, the AI computing module may include one or more artificial neural networks. Therefore, at least one AI-based arithmetic operation may be different from at least one deterministic arithmetic operation.
[0012] The training data of the AI computing module can be based at least on the measurement signal of the sensor device. This means in particular that the AI computing module that needs the training data for training during the operation of the field device receives the training data at least in the form of the measurement signal of the sensor device or based on the measurement signal of the sensor device. The measurement signal is detected by the sensor device and is processed or provided to the computing device without processing, so that it can use the measurement signal to determine the measurement value used to characterize or indicate the process parameter, such as the level of the container in the field device (in the form of a level measuring device). The training data may include or be based on these processed or unprocessed measurement signals. In particular, the measurement signal can be processed to obtain the training data. In addition to the collection, the processing can also include the cleaning, conversion and / or formatting of the measurement signal. In particular, the training data can be configured to train the AI computing module of the field device. In particular, the configuration of the training data can include the nature of the training data so that it can be used to train the AI computing module. The nature can be specified by collecting the training data during the operation of the field device. The nature of the collected training data can also be changed to make it available for training. For example, the nature can be predetermined by the format and / or type or kind of the training data. Advantageously, specific properties of the training data (i.e., a specific format, such as the data format or the formatting of the content of the training data itself) can be specified in order to optimize the training of the AI computing module.
[0013] The field device can be configured to perform a training operation in parallel with the deterministic computing operation using a deterministic computing module. Training data can be collected during this training operation. The AI computing module can also be trained using the training data during the training operation. Advantageously, the field device can therefore be used before the AI computing module is trained. At the same time, while the training operation is running, the measured values of the process parameters can be determined by the deterministic computing module of the computing device. Therefore, not only can the measurement signal be detected to generate training data, but the process parameters can also be detected based on the deterministic arithmetic operation of the deterministic computing module. In addition, the measured values determined by the deterministic computing module can be used for training data. In other words, the training data can be based at least on the measured values of the computing device (in particular the deterministic computing module) (in addition to the measurement signal of the sensor device or as a substitute for the measurement signal of the sensor device).
[0014] In particular, the field device may also be configured to perform an AI-based computing operation of the AI computing module after a start-up criterion based on the training operation is met. Thus, the training operation may be terminated when the training has been sufficiently or completely completed, which is indicated by the start-up criterion being met. For example, the start-up criterion may be at least one of a predefined match of at least one determined measurement value of a process parameter from the deterministic computing module and at least one determined measurement value of a process parameter from the AI computing module, a predefined operating time of the field device, and a predefined data volume and / or data quality of the training data. Thus, the start-up criterion may quantitatively and / or qualitatively indicate when the AI computing module is ready for an AI-based computing operation and may be "turned on" by the computing device.
[0015] The field device can be configured so that the execution of the AI-based computing operation ends the deterministic computing operation. As a result, the computing and / or energy resources of the computing device do not have to be additionally used for the deterministic computing operation, so that resource-saving operation of the field device is possible. In addition, after the start-up criteria are met, the AI-based computing operation can be regarded as more advantageous than the deterministic computing operation of the computing device. Finally, the AI computing module can be used to more reliably detect process parameters and / or additional functions (such as fault detection or parameterization changes), and / or implement them. The end of the deterministic computing operation can only be temporary, or, as will be described in detail below, at least the deterministic computing operation can be restored when the AI computing module is verified or the AI computing module is retrained.
[0016] Additionally or alternatively, the field device can be configured so that the deterministic computing operation and the AI-based computing operation are performed in parallel. This has the advantage that the measured values determined by the corresponding computing operations can be checked against each other to improve the accuracy of process variable detection.
[0017] Thus, the field device can be configured to determine the measured value based on a combination of a deterministic calculation operation with an AI-based calculation operation performed in parallel. In this case, for example, the average value of the measured values determined in each case can be used as the measured value determined by the field device and output, in particular displayed, if necessary. Alternatively, for example, based on a check of the two measured values, in particular using the measurement signal and / or the history of the measured values and / or by the AI calculation module, the deterministic calculation module and the AI calculation module can select one of the two measured values for the field device to determine and, if necessary, output.
[0018] Furthermore, the field device may be configured such that the AI computing module is verified at least based on the measured values determined by the deterministic computing module. This allows the AI computing module to be checked against the measured values determined by the deterministic computing module to ensure that the AI computing module is sufficiently trained and / or operates reliably with new data (i.e., measurement signals). The verification may also be performed to check whether the aforementioned start-up criteria are met, wherein the start-up criteria may in particular be a predefined match between at least one determined measured value of a process parameter from the deterministic computing module and at least one determined measured value of a process parameter from the AI computing module. Otherwise, if the verification shows that the result of the AI computing module using new data does not yet meet the requirements of the start-up criteria, the AI computing module may also be further optimized and further trained.
[0019] In addition, the field device can be configured to retrain the AI computing module when at least one parameter affecting the process parameter detection changes. In other words, if the previously explained parameterization changes, the field device can trigger the retraining of the AI computing module. A limit value can be selected for the change of the parameter affecting the process parameter detection, so that the AI computing module is retrained only when the parameter change that has occurred exceeds this limit value. As a result, the AI computing module can be prevented from being retrained when the parameterization changes slightly (for example, if the parameter is the ambient temperature and is only affected by normal fluctuations). Retraining can be carried out in the following way: the AI computing module (in particular, at least one neural network of the AI computing module) is partially or completely deleted and rewritten or supplemented with a partial or completely new AI computing module. Alternatively, the existing AI computing module can also be updated by retraining without deleting part or all of the AI computing module. For retraining, it is possible to switch to a deterministic computing operation in which the deterministic arithmetic operations of the deterministic computing module are performed.
[0020] The deterministic computing module can be implemented in a first computing unit. The AI computing module can be implemented in a second computing unit. These computing units are particularly physical units, in which the computing module is contained as a computer program product, and is particularly stored in at least one data memory of the computing unit and can be executed by the corresponding processor of the computing unit. The computing units can share a data memory, or each have a separate data memory storing one of the computing modules. The computing units can also share a processor at least temporarily or completely, or each have its own processor to execute the corresponding computing module and thereby perform the corresponding arithmetic operation or calculation operation. For example, the first computing unit can have a microcontroller, an x86 architecture, a von Neumann architecture, a Harvard architecture and / or a hybrid of a von Neumann architecture and a Harvard architecture. The second computing unit can have a neuromorphic processor, a neurosynaptic processor and / or a processor based on a neural device.
[0021] In principle, the sensor device can have a sensor circuit for detecting a measurement signal related to a process parameter and / or a specified sensor circuit. The sensor device can have one or more sensors, such as a level sensor, a radar-based level sensor, a limit level sensor, a flow sensor, a flow velocity sensor, a temperature sensor, an acceleration sensor and / or an actuator (e.g. a piezoelectric actuator), which can provide a measurement signal in the form of an input voltage, for example via a consumption voltage.
[0022] In particular, the sensor device may provide one or two measurement signals. In the case of two measurement signals, the measurement signals may be different from each other. In particular, the measurement signals may come from different sensors in the sensor device. Then, one of the two measurement signals may be used for the AI computing module described herein, and the other for the deterministic computing module described herein. Then, the AI computing module may be trained using one measurement signal and validated using the measurement signal for the deterministic computing module, in particular by supervised learning.
[0023] In particular, the sensor device can have a sensor and / or an actuator in the form of an acceleration sensor of a field device. The actuator can be, for example, a piezoelectric actuator, in particular a piezoelectric actuator of an electronic vibration limit level measuring device. For example, the actuator can provide a measurement signal in the form of an input voltage through its voltage consumption. Such measurement signals from the acceleration sensor and the actuator are related to process parameters of the corresponding level measuring device, in particular a vibration limit level measuring device, wherein the process parameter can be the reaching or not reaching of at least one or exactly one level when a part of the sensor device is covered by the medium. Such a limit level measuring device can also be called a limit level switch, because only two states can be distinguished here, namely the non-reaching and reaching of the predetermined level. Such a sensor device is particularly affected by the parameterization of the field device in a specific application and therefore particularly benefits from the training of the AI computing module during the operation of the field device in a specific application field. Ultimately, the computing device must determine whether the predetermined level has been reached, so that the detection of the process parameter has a decisive influence on the filling of a container in which the field device is contained, for example.
[0024] In principle, the field device can be used as any type of level measuring device, in particular as a radar-based level measuring device for detecting the level of a medium, as an electronic vibration limit level measuring device, as a flow measuring device for detecting the flow of a medium, as a flow rate measuring device for detecting the flow rate of a medium, as a pressure measuring device for detecting a pressure or differential pressure and / or as a temperature measuring device for detecting a temperature. The field device can also be any other field device.
[0025] However, in particular, the field device can be designed as an electronic vibration limit level measuring device, or in other words as a vibration limit level measuring device, in particular for monitoring the limit level by detecting the coverage of at least a part of the sensor device by the medium. Depending on the level in the measuring container, whether such a limit level measuring device is in contact with the medium, the oscillation frequency / damping or oscillation amplitude of the diaphragm of the limit level measuring device or the mechanical oscillator arranged on the diaphragm is affected by the contact with the medium. Such an electronic vibration limit level measuring device that detects the medium by detuning the oscillation system of the limit level measuring device can have at least one piezoelectric system for exciting and / or evaluating the oscillation system. When interacting with the medium, the oscillation element of the oscillation system (e.g. a tuning fork) can be detuned by interacting with the medium in order to detect the medium. Alternatively, at least one electromagnetic system can be used to excite and evaluate the oscillation system. Such a level measuring device is particularly affected by the parameterization of its settings and environment, and therefore particularly benefits from the training of the AI computing module during its operation in a specific operating environment.
[0026] The design of the electronic vibration limit level measuring device can be such that, for example, it has a mechanical oscillation system for detecting a medium. An actuator or in other words a driver can be used to excite the mechanical oscillation system. In order to detect the oscillation of the mechanical oscillation system, the level measuring device can have an acceleration sensor. The actuator or the driver can be mechanically coupled to the mechanical oscillation system. The mechanical oscillation system can have an oscillation element, and the limit level measuring device can be configured so that the oscillation element can interact with the medium so as to change the frequency and / or amplitude and / or damping of the mechanical oscillation system by interacting with the medium. The acceleration sensor can be mechanically coupled to the oscillation system, in particular to the diaphragm and / or the actuator or the driver of the oscillation system, in order to detect the oscillation of the mechanical oscillation system. The measurement signal (in particular the electrical measurement signal) generated by the acceleration sensor can be provided to a computing device in order to generate a measured value of a process parameter to be detected by the field device, whereby the measured value can depend on the interaction of the mechanical oscillation system with the medium. Alternatively or additionally, the voltage on the actuator (in particular the piezoelectric actuator) can also be used as a measurement signal, which can be detected by the voltage consumption of the actuator.
[0027] In particular, the measurement signals from the acceleration sensor and the actuator can be detected. The two measurement signals can be provided together and / or in parallel to the calculation device to determine the measurement value. In this way, a further improvement of the measurement can be provided.
[0028] Furthermore, not only the current measurement signal can be evaluated to determine the measurement value. The history of the measurement signal can also be used to determine the measurement value. The history of the measurement signal can be a record of previous measurement signals. Furthermore, this history of the measurement signal and / or the measurement value can also be used as a further basis for training the AI calculation module.
[0029] A second aspect of the invention relates to a method for training an AI computing module based on artificial intelligence on a field device for detecting a process parameter, the field device being particularly designed as a level measuring device for detecting the level of a medium, wherein the field device has a sensor device configured to detect a process parameter associated with a measurement signal, and wherein the field device has a computing device configured to determine a measured value of the process parameter based on the measurement signal, wherein the computing device has a deterministic computing module and an AI computing module, wherein the AI computing module is trained during operation of the field device.
[0030] A third aspect of the invention relates to a method for operating a field device for detecting a process parameter, the field device being particularly designed as a level measuring device for detecting a fill level of a medium, wherein the field device has a sensor device configured to detect a process parameter associated with a measurement signal, and wherein the field device has a computing device configured to determine a measured value of the process parameter based on the measurement signal, wherein the computing device has a deterministic computing module and an AI computing module based on artificial intelligence, the method comprising:
[0031] - training an AI computing module according to the method according to the second aspect of the invention,
[0032] - confirm that the training-based activation criteria are met, and
[0033] -Execute AI-based computing operations of the AI computing module.
[0034] The method according to the second or third aspect of the invention may be performed or may in particular be performed on the herein described field device according to the first aspect of the invention.
[0035] The method according to the second or third aspect of the present invention may also include any one or more of the following steps:
[0036] - performing a training operation using the deterministic computing module in parallel with the deterministic computing operation, wherein, in particular, training data is collected during the training operation and / or the AI computing module is trained using the training data,
[0037] - terminating a deterministic computing operation, in particular during or after execution of an AI-based computing operation,
[0038] -Execute deterministic and AI-based computational operations in parallel (in time),
[0039] - Determine the measurement value based on a combination of deterministic and AI-based computing operations performed in parallel,
[0040] - verifying the AI computing module based at least on the measurements determined by the deterministic computing module,
[0041] - Retraining the AI calculation module if at least one parameter affecting the detection of process measurements changes.
[0042] A fourth aspect of the present invention relates to a computer program product comprising instructions which, when the program is executed by a computer, cause the computer to perform the method according to the second aspect of the present invention or the third aspect of the present invention.
[0043] The computer program product according to the fourth aspect of the invention can in particular be executed or executable and / or stored on the field device described herein according to the first aspect of the invention. The computer program product can be a computer program device or software device, which comprises at least one computer program or at least one computer program code itself or a product, such as a memory on which such a computer program device or at least one computer program is stored.
[0044] A fifth aspect of the present invention relates to training data for training an AI computing module based on artificial intelligence on a field device for detecting a process parameter, the field device being particularly designed as a level measuring device for detecting the level of a medium, wherein the field device has a sensor device for detecting a process parameter associated with a measurement signal, and wherein the field device has a computing device configured to determine a measured value of the process parameter based on the measurement signal, wherein the computing device has a deterministic computing module and an AI computing module, wherein the training data is collected during operation of the field device.
[0045] The training data according to the fifth aspect of the invention can be used in particular on the field device described herein according to the first aspect of the invention and / or collected from the field device. The training data can be used specifically for training the AI computing module of the field device. The device can include in particular the quality of the training data so that it can be used to train the AI computing module. The condition can be specified by collecting the training data during the operation of the field device. The properties of the collected training data can also be changed to make it available for training. For example, the property can be predetermined by the format and / or type of the training data. Advantageously, specific properties of the training data (i.e., a specific format, such as the data format or the formatting of the training data content) can be specified in order to optimize the training of the AI computing module.
[0046] Features, elements, functions and / or advantages of the field devices described herein may also be applied to the methods, computer program products and training data described herein, and vice versa, and may be arbitrarily combined with one another.
[0047] Embodiments of the present invention will be described below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 A block diagram of a field device according to an embodiment is schematically shown.
[0049] Figure 2 A flow chart illustrating the steps of a method for operating a field device according to an embodiment is schematically shown.
[0050] Figure 3 Schematically shows a schematic diagram for further explanation according to an embodiment Figure 2Another flow chart of the method.
[0051] Figure 4 Schematically shows a circuit according to another embodiment for further explanation Figure 2 Another flow chart of the method.
[0052] Figure 5 A schematic diagram is shown for further explanation according to another embodiment. Figure 2 Another flow chart of the method.
[0053] Similar, similarly acting, identical or identically acting elements may be provided with similar or identical reference symbols. DETAILED DESCRIPTION
[0054] Figure 1 A block diagram of a field device 10 according to an embodiment is schematically shown. Figure 1 The field device 10 is designed as an example of a level measuring device 10 . Alternatively, the field device 10 may be a pressure measuring device, a temperature measuring device, a flow rate measuring device, a flow rate measuring device or any other field device 10 .
[0055] The field device 10 is configured to detect a process parameter, in particular the level of a medium in a container (not shown). In particular, the field device 10 can be configured as an electronic vibration limit level measuring device, which only determines whether the medium in the container has reached a predetermined level or a limit level.
[0056] The field device 10 comprises a sensor device 12. The sensor device 12 comprises at least one sensor 14, which is designed here, for example, as an acceleration sensor of an electronic vibrating limit level measuring device. The acceleration sensor measures acceleration, which is detected here as a measurement signal by the sensor device 12. In addition, the sensor device 12 comprises an actuator 16, in particular a piezoelectric actuator, whose voltage consumption can be detected as a measurement signal that is correlated with a process parameter.
[0057] In principle, however, the field device 10 can also comprise only one sensor 14 or one unit for detecting a measurement signal, in particular according to one measuring principle. The method explained in more detail later can or at least only evaluate one such measurement signal, which is based on only one measuring principle or measuring method, or alternatively can evaluate a plurality of measurement signals and in particular detect or determine a process parameter from them.
[0058] The field device 10 further comprises a computing device 20. The computing device 20 may be configured to control the sensor device 12. In particular, the computing device 20 is configured to process measurement signals detected by the sensor device 12 and / or to determine a measured value of a process parameter.
[0059] The computing device 20 includes a deterministic computing module 22. The computing device 20 also includes an AI computing module 26 based on artificial intelligence (AI). The AI computing module 26 may include at least one artificial neural network. The deterministic computing module 22 can be executed by the computing unit 24 of the computing device 20, in particular in the form of a computer program, and / or can be stored in the first computing unit 24, in particular in a data memory included in the first computing unit 24. The deterministic computing module 22 can be executed by a processor (not shown) of the first computing unit 24. The AI computing module 26 can be executed by the second computing unit 28 of the computing device 20, in particular in the form of a computer program, and / or can be stored in the second computing unit 28, in particular in a data memory included in the second computing unit 28. The AI computing module 26 can be executed by a processor (not shown) of the second computing unit 28.
[0060] Therefore, the first and second computing units 24, 28 can be designed and / or arranged in separate housings individually, separately and / or separately. Alternatively, the two computing units 24, 28 can also be arranged in a common housing. Alternatively, the deterministic computing module 22 and the AI computing module 26 can be executed by a single processor and / or stored on a common data memory. In particular, the first and second computing units 24, 28 can have a common storage area or a separate storage area in a data memory (not shown) of the field device 10.
[0061] The two computing modules 22, 26 and / or the two computing units 24, 28 are connected to each other via a communication link 29. Thus, communications (in particular data) can be exchanged between the two computing modules 22, 26 and / or the two computing units 24, 28. As will be explained in more detail below, this is useful, for example, for training of the AI computing module 26, for verification of measured values of the two computing modules 22, 26, and for many other applications.
[0062] The deterministic calculation module 22 is designed for a deterministic calculation operation 23, which can include one or more deterministic arithmetic operations. Such arithmetic operations can be associated, for example, with the processing of the detected measurement signal (in particular, the determination of a measured value of a process parameter based on the measurement signal, the scaling for determining the measured value based on the measurement signal, the unit conversion for determining the measured value based on the measurement signal) and / or the monitoring of the function of the field device 10.
[0063] The AI calculation module 26 is designed to perform an AI calculation operation 27, which may include one or more non-deterministic, probability-based and / or classification-based arithmetic operations. Such arithmetic operations may be associated with features that have been explained for the arithmetic operations of the deterministic calculation module 22. In addition, the arithmetic operations may be associated with aging detection of at least a portion of the field device 10, drift determination of the sensor device 12, contamination determination of at least a portion of the sensor device 12, signal quality determination of the measurement signal, signal strength determination of the measurement signal, detection of at least one interference influence on the measurement signal and / or electromagnetic interference detection of at least a portion of the field device 10.
[0064] The first computing unit 24, the second computing unit 28 and / or the two computing units 24, 28 (in particular the entire computing device 20) and at least the sensor 14 can also form an integrated system, in particular connected to each other, arranged adjacent to each other, and / or arranged in a common housing. Such an integrated system or any computing unit 24, 28 or computing device 20 can be arranged close to the corresponding piezoelectric element of the actuator 16 or directly on the corresponding piezoelectric element of the actuator 16. In particular, close can mean that there is only one element between the corresponding components, in particular only one circuit board. This allows a compact design of the field device 10 to be achieved.
[0065] Finally, the field device 10 may also comprise an output unit 30. The output unit 30 may be, for example, a display. The output unit 30 may output, in particular display, a determined measured value of a process parameter and / or a process parameter.
[0066] Figure 2 Schematically illustrates an operation for Figure 1 Flow chart of steps S1, S2, S3 of the method of the field device 10 in FIG. In the first step S1, the AI calculation module 26 is trained. The first step S1 is also referred to as the training operation 25 (see Figures 3 to 5 ). In a second step S2, it is checked or determined whether the training-based activation criteria have been met. In a third step S3, the previously explained AI-based calculation operation 27 of the AI calculation module 26 is performed.
[0067] In accordance with Figures 3 to 5 In the schematic diagram of the method of various embodiments in FIG. 1 , steps S1 to S3 are shown in an exemplary time sequence.
[0068] Figure 3The start time S is shown when the operation of the field device 10 starts. Starting from this start time S, the operation of the field device 10 starts, which includes a deterministic calculation operation 23, with which the calculation device 20 can determine the measured value of the process parameter based on the measurement signal of the sensor device and thus can detect the process parameter. At the same time, a training operation 25 is performed.
[0069] Starting from a start time E in the runtime or operating time of the operation of the field device 10, which occurs when the start criterion according to the second step S2 is met, the AI calculation module 26 is fully or completely trained and can now be used. The start criterion can be, for example, at least one of a predefined match of at least one determined measured value of a process parameter from the deterministic calculation module 22 and at least one determined measured value of a process parameter from the AI calculation module 26, a predefined operating time of the field device 10, and a predefined data volume and / or data quality of the training data. Thus, the deterministic calculation operation 23 ends at the start time E, and the AI calculation operation 27 is "started" or executed, which corresponds to the third step S3.
[0070] Figure 4 Shows Figure 3 A variation of the method in which the deterministic calculation operation 23 is not terminated at the start time E, but continues to run in parallel with the AI calculation operation 27.
[0071] The determination of the measured value after the switch-on time E can be performed based on a combination of deterministic calculation operations 24 and AI-based calculation operations 27 performed in parallel.
[0072] The AI computing module 26 may also be verified based at least on the measured values determined by the deterministic computing module 22. As an example, a verification time V is shown here when verification may be performed. Of course, multiple verification times V may be specified or exist, and continuous or continuous verification is also possible.
[0073] Figure 5 Shows Figure 3An extension of the method of , in which the parameterization of the field device 10 is changed at a parameter change time P. The parameterization can affect the field device 10 itself (in particular its settings) as well as the environment (in particular the environmental conditions outside or inside the container, such as the density of the medium). The previous AI calculation operation 27 is then ended at the parameter change time P and the deterministic calculation operation 23 is used again. At the same time, a new training operation 25 is started to train a new AI calculation module 26 or to update an existing AI calculation module 26 by retraining, and then the AI calculation module 26 is used again at the power-on time E to perform a new AI calculation operation 27. The field device 10 can thus react particularly flexibly to parameter changes in order to particularly ensure a high measurement accuracy and / or a low susceptibility to errors.
[0074] In addition, it should be noted that "comprising" and "having" do not exclude other features or steps, and the indefinite article "a" or "an" does not exclude a plurality. Furthermore, it should be noted that features or steps that have been described with reference to one of the above embodiments can also be used in combination with other features or steps of other above-described embodiments. Reference signs in the claims should not be considered as limiting.
[0075] REFERENCE TO RELATED APPLICATIONS
[0076] This application claims the priority of German patent application 10 2023 114724.2 filed on June 5, 2023, the entire contents of which are incorporated herein by reference.
Claims
1. A field device (10) for detecting a process parameter, which is particularly designed as a level measuring device for detecting the level of a medium, wherein: The field device (10) comprises: - a sensor device (12) configured to detect a measurement signal related to the process parameter, and - a computing device (20) configured to determine a measured value of the process parameter based on the measurement signal, wherein the computing device (20) comprises a deterministic computing module (22) and an AI computing module (26) based on artificial intelligence (AI), Wherein the field device (10) is configured to train the AI computing module (26) during its operation.
2. The field device (10) according to claim 1, wherein: The training data of the AI calculation module (26) is based at least on the measurement signal of the sensor device (12).
3. The field device (10) according to claim 1 or 2, wherein: The field device (10) is configured to use the deterministic computing module (22) to perform a training operation (25) in parallel with a deterministic computing operation (23), in which training data is collected and the AI computing module (26) is trained using the training data.
4. The field device (10) according to claim 3, wherein: The field device (10) is configured to perform an AI-based computing operation (27) of the AI computing module (26) after a start-up criterion based on the training operation is met.
5. The field device (10) according to claim 4, wherein: The field device (10) is configured such that the execution of the AI-based computing operation (27) ends the deterministic computing operation (23), and / or the deterministic computing operation (23) and the AI-based computing operation (27) are executed in parallel.
6. The field device (10) according to claim 5, wherein: The field device (10) is configured to determine the measurement value based on a combination of the deterministic computing operation (23) and the AI-based computing operation (27) performed in parallel therewith.
7. The field device (10) according to any one of the preceding claims, wherein: The field device (10) is configured such that the AI calculation module (26) is verified based at least on the measured value determined by the deterministic calculation module (22).
8. The field device (10) according to any one of the preceding claims, wherein: The field device (10) is configured to retrain the AI computing module (26) when at least one parameter affecting detection of the process parameter changes.
9. The field device (10) according to any one of the preceding claims, wherein: The deterministic computing module (22) is implemented in a first computing unit (24), and the AI computing module (26) is implemented in a second computing unit (28).
10. The field device (10) according to any one of the preceding claims, wherein: The sensor arrangement (12) comprises a sensor (14) and / or an actuator (16) of the field device (10), the sensor being in the form of an acceleration sensor.
11. The field device (10) according to claim 10, wherein: The field device (10) is designed as an electronic vibrator limit level measuring device.
12. A method for training an AI computing module (26) based on artificial intelligence (AI) on a field device (10) for detecting a process parameter, the field device (10) being particularly designed as a level measuring device for detecting a level of a medium, wherein: The field device (10) comprises a sensor device (12) configured to detect a measurement signal related to the process parameter, and wherein the field device (10) comprises a computing device (20) configured to determine a measured value of the process parameter based on the measurement signal, wherein the computing device (20) comprises a deterministic computing module (22) and the AI computing module (26), Wherein the AI computing module (26) is trained during operation of the field device (10).
13. A method for operating a field device (10) for detecting a process parameter, the field device (10) being designed in particular as a level measuring device for detecting a fill level of a medium, wherein: The field device (10) comprises a sensor device (12) configured to detect a measurement signal related to the process parameter, and wherein the field device (10) comprises a computing device (20) configured to determine a measured value of the process parameter based on the measurement signal, wherein the computing device (20) comprises a deterministic computing module (22) and an AI computing module (26) based on artificial intelligence (AI), the method comprising: - training the AI computing module (26) according to the method according to claim 12; - confirming that training-based activation criteria are met; and - Execute AI-based computing operations (27) of the AI computing module (26).
14. A computer program product comprising instructions which, when said program is executed by a computer, cause said computer to perform the method according to claim 12 or 13.
15. Training data for training an AI computing module (26) based on artificial intelligence (AI) on a field device (10) for detecting a process parameter, the field device (10) being particularly designed as a level measuring device for detecting a level of a medium, wherein: The field device (10) comprises a sensor device (12) configured to detect a measurement signal related to the process parameter, and wherein the field device (10) comprises a computing device (20) configured to determine a measured value of the process parameter based on the measurement signal, wherein the computing device (20) comprises a deterministic computing module (22) and the AI computing module (26), Therein, the training data has been collected during operation of the field device (10).
