Parameter setting of components in an automated system

Abstract

The present invention relates, in one aspect, to a method for basic parameter setting of a component (K) in a test automation system (AA). The method comprises: starting a test run (S2) of the component (K) in the automation system (AA) with basic parameter setting; measuring (S3) measured value data records during the test run; accessing a machine learning module (ML) including a pre-trained neural network (ANN), wherein the pre-trained neural network (ANN) is pre-trained to calculate (S4) a target parameter setting for the corresponding component (K) for a measured value data set, and wherein the basic parameter setting is compared with the calculated target parameter setting, and in the event of a deviation, providing (S5) a result message for adjusting the basic parameter setting; receiving the provided result message for adjustment of the basic parameter setting (S6).

Description

Technical Field

[0001] The present invention relates to the parameter setting of components (such as electric or pneumatic actuators) in an automation system before the components are put into operation. Background Art

[0002] The commissioning of components in automation technology usually requires parameter setting of the components to achieve adjustment for the intended purpose. Thereby, the components are adapted to a specific problem definition, such as the mass to be moved, the travel speed, etc. In particular, the mass to be moved must be parameterized, for example, using a controlled drive. This requires certain expertise on the customer side. Therefore, various support services are currently provided to facilitate commissioning. For example, special design tools based on the customer's basic application parameters provide customized system configurations and parameter sets before commissioning. Incorrect configuration of control parameters may lead to machine failures, which may damage the equipment.

[0003] For example, it is known in the prior art that there is a configuration software (such as Festo Automation-Suite), which provides a commissioning wizard for a special automation system (i.e., an electric drive). The commissioning wizard guides the user through the parameter setting of the component and provides assistance. However, the customer has to determine and enter all values by himself. For electric drives, the "auto-tuning" mode is used, which can determine certain system parameters. The disadvantage of this mode is that the customer has to manually determine and transmit all the required data locally. If too little data is selected or the wrong data is selected here, the help may be insufficient or not helpful at all. In addition, the data or information obtained locally at a specific device cannot be used globally and thus can sometimes also be used for other devices. This is where the present invention comes into play.

[0004] The problem with the previous help is that the expertise of the device operator is always required and thus a manual method or at least a manual check is required. Summary of the Invention

[0005] Based on the above situation, the object of the present application is to better support customers before and during the commissioning of components in an automation device. In particular, automatic parameter setting of the components in the corresponding application environment should be possible. In addition, it should be possible to provide parameter setting for all types of systems, for example, not only for electric drives but also for pneumatic systems. In addition, the method should not depend on the selection and (possibly manual) setting of customer data.

[0006] This task is solved respectively by the objects according to the independent claims, in particular by a method for checking basic parameter setting, a parameter setting unit, and a component having such a parameter setting unit. Advantageous designs are the subject matter of the dependent claims, the description, and the drawings.

[0007] In a first aspect, the present invention relates to a computer-implemented process that is performed locally on-site and, in particular, directly on components (i.e., in an automation system), and is intended for testing the basic parameter settings of components in an automation system (especially in factory settings). The following process steps are performed before or during the commissioning of components in an automation device:

[0008] - Start a test run of the components in the automation system using the basic parameter settings;

[0009] - Measure a dataset of measured values during the test run (using the basic parameter settings);

[0010] - Access a machine learning module that may include pattern recognition algorithms and / or pre-trained neural networks, where the pre-trained neural network (hereinafter also referred to as "artificial neural network" ANN) is pre-trained to calculate the target parameter settings for the corresponding components for the measured value data record, and where the basic parameter settings are compared with the calculated target parameter settings, and in the event of a deviation, a result message for adjusting the basic parameter settings is provided;

[0011] - Receive the provided result message to adjust the basic parameter settings.

[0012] The advantage of the method described herein is that commissioning can be further automated and simplified. Based on the signals recorded during commissioning that represent the specific and real operation of the components in the device, suggestions for optimizing the parameters used can be calculated and output specifically for the corresponding use of the components in the device. Another advantage of the method described herein is that it can be applied independently of the device type.

[0013] A preferred application of the present invention is in the pneumatic field, in particular for adjusting the basic parameter settings of controlled pneumatic valves. For this purpose, application parameter settings are required, which can be verified based on the suggestions. Servo-pneumatic requires additional adjustments in addition to the control of electric servo drives because the control algorithm is much more complex due to the compressible medium air. In a preferred training, the method for a pneumatic system can be optimized, for example, to minimize air consumption, and suggestions for the optimal design of the components can be provided.

[0014] In a preferred form of the present invention, the machine learning module is continuously retrained using the continuously recorded dataset of measured values, the basic parameter settings, and the calculated target parameter settings. The machine learning module thus stores a self-learning model that is provided with continuously newly measured data and continuously "keeps learning". This model can advantageously be used for other systems and / or for creating basic parameter settings.

[0015] In another preferred embodiment of the present invention, it is possible to propose to provide and pre-train specific machine learning modules for individual selection components, especially for complex components. Thus, in the configuration phase, the components for which, for example, a specific ANN is to be trained can be selected and determined. Thus, the method can provide, for example, basic parameter settings for simple components and / or specifically adjusted parameter settings for selected components (such as control nodes, such as SPS / PLC) and / or optimization of basic parameter settings.

[0016] In an advantageous further training of the present invention, the ANN and / or the machine learning module is trained or pre-trained using laboratory data and / or data from a test bench and / or data from at least one simulation model. Thus, a larger database can be used for the pre-training of the network, in particular, test data from component and / or device manufacturers can be used. Customers who want to operate components and devices usually do not have a large amount of data during commissioning. Therefore, improved and specific parameter settings can be provided at an earlier time point.

[0017] In another advantageous further training of the present invention, the machine learning module includes a pattern recognition algorithm that is trained to automatically recognize error patterns in basic parameter settings, especially errors or unfavorable parameter settings for the specific use of components in a device for a corresponding automation task. For example, this includes conclusions drawn from the detected oscillation behavior, time / speed to reach the end position, etc. The recorded measured values (actual data set) can be analyzed relative to reference values (nominal data set) that may be pre-measured in the laboratory, especially by a pattern recognition algorithm (which will be described in more detail below), in order to draw conclusions about error patterns based on this analysis. The model can be designed to distinguish between good patterns and error patterns. In a preferred design, a set of measured signal characteristics is standardized and certain patterns are identified therein. Thus, a pure threshold analysis is not performed, but preferably pattern recognition is performed on several time-related signals.

[0018] In an advantageous embodiment of the present invention, the machine learning module can be adapted to apply at least one pattern recognition algorithm to a data set of measured values. The pattern recognition algorithm can be applied, for example, to the measured measured values and can be designed to distinguish between good cases and error cases. Examples of specific patterns that can be purposefully identified are: the trend of system oscillation, traveling too "hard" to the end position, or tracking errors due to incorrect interpretation of application parameters.

[0019] In another advantageous embodiment of the invention, the machine learning module may include a component testing unit that is trained to test whether the correct or appropriate components are fully used in an automated system to complete a given automated task, and in the negative case, the component testing unit extends the result message with a swap message (for swapping components). The swap message may also include a set of improved parameters that replace the set of parameters when needed and / or especially in response to an acknowledgement signal (basic parameter setting -> improved parameter setting) that may be input on a component or control unit of the device. Preferably, the component testing unit includes a parameter analysis unit that, as described above, is designed to detect certain problems in the parameter setting. For example, it may be detected that the torque limit is too low. Then, the component testing unit may recommend using a larger electric motor.

[0020] In another advantageous embodiment of the invention, the machine learning module may be formed at least partly locally on a component or on a selected component. The advantage of this is that each (each selected) component can act and make decisions autonomously. Thus, the component can autonomously set its own parameters and rewrite the basic parameter setting such as the factory settings. However, this requires the local component to have a high computing power. The local component can also be designed as an embedded system or include such a system. The component can be designed with or without special hardware acceleration.

[0021] In an alternative embodiment of the invention, the machine learning module may also be trained centrally on a server, for example. The server may be, for example, the server of a component manufacturer. The server can be trained in the cloud. The advantage of this is that the knowledge collected and learned about a specific system can be made available to other system operators in an anonymous or more abstract form.

[0022] The method can also be implemented as a distributed system. Thus, the first method step is implemented and executed on a first unit, and the second method step (distributed) is implemented and executed on a second unit. Preferably, the evaluation (inference) and learning (from samples) in the debugging tool (local, on the device) should be executed on a central server (e.g., of the manufacturer of the component / system). Thus, the evaluation can be executed locally on the component. Then, an appropriate diagnostic message is sent to inform the operator about the poor parameter setting. An advantageous further training of the invention provides a preprocessing of the data on the component. Data preparation (such as filtering or averaging) is especially required for high-frequency signals such as the motor current. It is preferred to evaluate the signal in the frequency domain, i.e., evaluate the spectrum of the signal, for example by applying the fast Fourier transform. The evaluation of the component also requires intermediate storage of the signal over a certain period of time. Thus, the local component preferably has a data storage device.

[0023] Preferably, existing configurations and engineering interfaces are used. There is no need to implement new Internet of Things (IoT) or data collection interfaces. This can greatly reduce the installation workload.

[0024] In addition, in another preferred embodiment of the present invention, a special test sequence is run for debugging to obtain relevant data. In this case, it is clearly stated that the data collected during this process is not operational data, so this data collection is different from the data collection during operation.

[0025] The measured values recorded or measured are basically not limited to a certain type of measured value or type of measured value. For example, the measured value data set can include digital and / or analog signals and / or signal characteristics that change over time. The measured values can be obtained by different sensors (position sensors, end position sensors, speed or temperature sensors, or other types of sensors) within an automated system. The sensors do not necessarily have to be directly mounted on the components.

[0026] If the computer program is executed on a computer or a computer-based electronic instance (such as a microcontroller, PLC), then another object solution is a computer program that has computer program code for performing all the method steps of the above-described method. The computer program can also be stored on a computer-readable medium.

[0027] The object solutions have been described above using the method. The features, advantages, or alternative designs mentioned can be similarly transferred to other claimed subject matters, and vice versa. In other words, apparatus claims (such as claims directed to a parameter setting unit or a component) can also be further improved by using the features described or claimed in connection with the method. Thus, the corresponding functional features of the method are formed by the corresponding device modules of the apparatus, in particular by hardware modules or microprocessor modules, and vice versa. To avoid redundancy, these alternative embodiments of the apparatus claims are not explicitly repeated here.

[0028] On the other hand, the present invention relates to a parameter setting unit for a component of an automated system, which is adapted to perform the method as described above. The parameter setting unit includes:

[0029] - a parameter interface for reading basic parameter settings;

[0030] - a sensor for measuring data or signals and / or at least one measured value interface for reading a measured value data set. This means that the parameter setting unit does not necessarily have to include a sensor for data acquisition itself, but alternatively or additionally only needs to include an interface through which the corresponding sensor data signal (measured value) is read. For example, the sensor can be located on another component in the automated system.

[0031] - An interface of the machine learning module through which result messages can be read to match the basic parameter setting with the target parameter setting.

[0032] - A processor for controlling and operating components through basic parameter setting and / or target parameter setting.

[0033] On the other hand, the present invention relates to a component in an automation system having such a parameter setting unit. The component may include a control unit and in particular a controller for an electrical or pneumatic system. The automation system may also include electrical and / or pneumatic actuators.

[0034] Hereinafter, the terms used in this application will be defined in more detail.

[0035] A "measurement data set" is a digital data set that can be electronically processed and is based on measured physical quantities or values. The measured values are recorded by different sensors and may refer to different physical quantities (such as temperature, position, pressure curve, or other time-varying signal curves, etc.). The sensors may be directly located on the component to be tested or may be located at other positions in the automation system, for example, on a component (such as a controller) that interacts with and / or is controlled by the component.

[0036] The automation system may include pneumatic and / or electric actuators. The automation system may be an electronic control system with various physical or technical components for different purposes, such as production equipment or a production line or a machine or a group of machines.

[0037] These components are technical parts or field devices that can be electronically controlled. The components in turn may have different types of component parts, such as analog components (valves, switches, etc.) and digital components (for example, software-based control units such as PLCs, etc.). According to the functions of the entire device and the automation tasks, the components are interconnected according to the circuit diagram to form functional connections. In this way, for example, an effect chain can be formed by several components connected in series. However, more complex structures (comprehensive ring-shaped or mesh-like component structures) can also be formed. The components may include sensors. They may also be designed with interfaces to read sensor signals and / or measurement values from other components and parts.

[0038] The machine learning module is an electronic unit that can be trained in software and / or hardware. The machine learning module may include a form of pattern recognition algorithm that is applied to the acquired measurement data. The machine learning module may include a pre-trained or trained neural network and, if necessary, may also include other machine learning methods. The machine learning module can be trained on a server and can be connected to a database.

[0039] A trained neural network (also known as an ANN (Artificial Neural Network)) is a computer-implemented method for calculating an optimized parameter dataset (target parameter setting). The ANN is based on training data obtained from laboratory and / or simulation data. The training data includes input data and output data. The input data includes measurement value data obtained from a component respectively. The output data includes the "optimal" parameter setting. The network is trained based on this training data so that it calculates the target parameter setting for any measurement value data of the component. A supervised training method can preferably be used for this training method. Further training also provides unsupervised training. A one-dimensional convolutional neural network is preferably used for the analysis of a time-limited signal curve (course). For detailed information on implementation, please refer to: Kiranyaz, Serkan and Avci, Onur and Abdeljaber, Osama and Ince, Turker and Gabbouj, Moncef and Inman, Daniel, (2019), 1D Convolutional Neural Networks and Applications (One-dimensional Convolutional Neural Networks and Their Applications). As an alternative method, a recurrent neural network (RNN) or long short-term memory (LSTM) and its variants can be used.

[0040] The basic parameter setting is a set of parameters preset at the factory. Therefore, a certain standard parameter setting can be provided with the component, but it has not been designed for a specific automation task. If this component is then operated in the device during the commissioning run, the measurement values are recorded and analyzed during the commissioning run. These measurement values are compared with the reference values of the component with the optimal parameter setting to provide the target parameter setting as a result of the machine learning process.

[0041] In the simplest case, the result message can be a signal. Alternatively, the result message can be a message packet that signals that the existing parameter setting of the component must be adjusted or optimized. Preferably, the result message contains the target parameter setting by which the existing parameter setting (basic parameter setting or change) will be rewritten. For this purpose, the result message can be displayed on the user interface with a request for an input to trigger the confirmation signal for rewriting.

[0042] The parameter setting unit is an electronic instance. It is used to execute the parameter setting method. Description of the Drawings

[0043] In the following detailed description of the drawings, non-limiting design examples and their features and other advantages are discussed based on the drawings. Shown in this figure are:

[0044] Figure 1 is a block diagram of a device according to a preferred embodiment of the present invention;

[0045] Figure 2 is a flowchart of a process according to a preferred embodiment of the present invention;

[0046] Figure 3 is a sequence diagram for signal exchange between participating instances according to a preferred embodiment of the present invention; and

[0047] Figure 4 is a schematic block diagram of a component having other components according to another preferred embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] In the following detailed description with reference to the drawings, non-limiting design examples and their features and other advantages are discussed based on the drawings.

[0049] The present invention proposes a feedback mechanism that, based on artificial intelligence algorithms and on previous experience, particularly with respect to components, calculates suggestions for improving previous or current parameter settings (component parameter settings, application parameter settings). Patterns in sensor data are recognized by pattern recognition algorithms to generate recommendations without exact knowledge of the complete physical structure of the system. These algorithms are implemented in a machine learning module and enable customers to access the expertise of component manufacturers regarding components and applications in order to optimize application and component parameters. In this process, previously known patterns in sensor data are recognized and, based on this, suggestions are transmitted to the user in the form of result messages. The idea here is to enable the customer to use comprehensive expertise from component development and testing via problematic patterns or degrees of significance in sensor data. A pre-trained model library can be provided.

[0050] The training of machine learning algorithms can be based on existing (e.g., from continuous operation) measurement values or on signal characteristics generated by simulation models. Using these methods, the basic parameter settings of components can be verified on-site. In addition, component selection can be checked and suggestions can be made if necessary. For example, the commissioning of a motor controller for a permanent magnet excited synchronous motor servo motor (PMSM) is described below.

[0051] For commissioning, many specific application parameters must be entered. This can be done via commissioning software. A set of measurement signals can be recorded to verify the parameter settings. This set of measurement signals preferably includes at least the following variables: regulated motor current; (DC bus) voltage; speed; speed change and acceleration; temperature; and other derived variables. In addition, setpoint specifications are also recorded. These signals are used as input (features) for the analysis function. Certain predefined (e.g., undesired) patterns, such as oscillation trends, are thereby recognized. Corresponding suggestions can then be given.

[0052] By using AI algorithms, it is possible to make instructions regarding parameter settings and proper use of components without knowing the exact physical model of the application.

[0053] Figure 1 A schematic illustration of an automation system AA (also simply referred to as system AA hereinafter), especially a system with pneumatic actuators, is shown. This automation system exchanges data with a server S at least temporarily, especially during the commissioning of a component K. The server S can be a computer instance installed within the domain of the device AA or preferably outside its domain. The server S can be operated, for example, by the component manufacturer or by the manufacturer of the AA system.

[0054] A machine learning module ML is implemented on the server S. This machine learning module can include a neural network ANN. The machine learning module ML can be trained to generate a model for the parameter settings of the component K. The generated model can be stored in a database DB.

[0055] As Figure 1 schematically shown, there are different design variants on one side of the AA system. Basically, the AA system includes a large number of different components K, such as various field devices and control devices (PLCs, etc.). As exemplified by components K1 and K2 Figure 1 shown, all or selected components K can include a parameter setting unit P. The component can be directly designed with a sensor Se to obtain measurement values on the component K. However, the component K can also be designed, for example, to have very few computer resources so that its own local parameter setting unit P cannot be designed. In this case, the measurement values of the component K n (in Figure 1 ) can be transmitted to a superior control unit ST, which thus includes the parameter setting unit P and performs parameter setting checks on behalf of the corresponding component K n Alternatively, another parameter setting unit P that is not assigned to the corresponding component or not installed on the corresponding component can also perform parameter setting checks on "foreign" components. Also as Figure 1 shown, the parameter setting unit P can not be directly implemented on the component K, but can be assigned to K as a separate component or as a separate instance of a component, and this parameter setting unit P has a corresponding data connection for data exchange.

[0056] The basic parameter settings can be provided by the factory when the component K is delivered and can thus be used as a data record on the component K, or can be read from the central server S via a data connection (preferably a wireless connection, such as radio). The measured values measured on or near the component K are transmitted in the form of a measured value data record to the machine learning module ML, which then uses this data to access the trained network ANN to calculate the target parameter settings. This can be, for example, in response to a signal transmitted from the server S to the component K, so that the parameter settings can be locally adjusted and optimized on the component K for the corresponding automation task.

[0057] Figure 2 The flowchart of a preferred embodiment of the present invention is shown. After starting the processing of the component K, the basic parameter settings are read in step S1. This can be done via the memory on the component K or via a data connection that can access the central server S. Then, the system AA operates in a test mode using the basic parameter settings. In step S3, the measured values are measured and / or recorded. Here, for example, it is recorded when certain actuators reach their respective final positions, how the pressure curve is, etc. Samples are extracted from the recorded measured values or signals and the samples are examined, the recorded measured values or signals such as: a very sudden stop in the final position of a pneumatic actuator by position measurement, or the controlled signal tends to oscillate, or an excessive deviation of the setpoint value / actual value in the controlled signal, or an unexpected force curve in a pressing application in some cases, or an unexpectedly high frictional force is generated in the case of incorrect mechanical design, which may indicate incorrect component dimensions.

[0058] In step S4, the pre-trained neural network ANN is accessed using the measured values to determine the target parameter settings. Thus, the machine learning model pre-trained for the corresponding component K analyzes the acquired sensor data or measured values and tries to identify certain patterns indicating optimization potential. Based on the detected anomalies, suggestions are displayed to the user (the operator of the AA machine with the corresponding component K) in the form of a result message. Based on the detected results, suggestions are displayed to the user to optimize the parameters. If the target parameter settings are different from the basic parameter settings, other steps can be initiated by accessing a set of rules (which can be stored, for example, in a rule base). For example, if the deviation is considered relevant, a result message can be created that includes a set of commands that initiate or indicate an adjustment to the basic parameter settings of the component K. Thereafter, the method can end or be repeated after a predetermined time unit. For example, the process can be repeated to check the parameter changes performed.

[0059] It is also possible to store new rewritten rules that trigger the previous parameter settings (e.g., an excessive deviation of the measured values or after a "restart" of the component and / or system, etc.).Figure 2 A vertical dashed line is drawn in Figure 2 . This line indicates that the steps to the left of the line can be performed on component K, and the steps to the right of the line can be performed on the machine learning module ML. Alternatively, all steps can be performed on component K, or the machine learning module ML can be directly installed on component K.

[0060] Figure 3 An interactive diagram shows which data is exchanged between the various components. The basic parameter settings are usually received and read by the server S. During the test run, the measured values are measured and read or recorded and transmitted to the network ANN to calculate the target parameter settings, which are then transmitted to component K. Then, the previous parameter settings (such as the basic parameter settings) can be overwritten to start the production operation of device AA.

[0061] Figure 4 Component K is shown in block diagram form and is designed to have a parameter setting unit P. In this design example, the parameter setting unit P includes three interfaces: a parameter setting interface PS, which is intended to read the basic parameter settings; a measured value interface MS, through which the measured values from sensors Se that are not arranged in or on component K are measured and read; and a learning module interface LMS of the machine learning module ML, so that the target parameter settings or result messages are read via the learning module interface LMS. In this implementation example, the learning module ML is directly implemented in component K; alternatively, it can also be implemented on an external server and / or implemented in the cloud and can be accessed via a corresponding network connection. The parameter setting unit P can also include local sensors or transducers and can be equipped with an electronic processing unit CPU.

[0062] In summary, it should be noted that the description and example embodiments of the present invention should not be understood as a limitation of a certain physical implementation of the present invention. All features explained and shown in connection with the various embodiments of the present invention can be provided in different combinations in the subject matter of the present invention to simultaneously achieve its advantageous effects.

[0063] The scope of protection of the present invention is given by the appended claims and is not limited by the features explained in the description or shown in the drawings.

[0064] In particular, it is obvious to those skilled in the art that the present invention can be applied not only to pneumatic actuators and components, but also to other types of automation devices (electrical devices). In addition, the components and / or parts of the parameter setting unit can be implemented as being distributed over several physical products.

[0065] Reference signs:

[0066] AA Automation Equipment

[0067] K Component

[0068] P Parameter Setting Module

[0069] ML Machine Learning Module

[0070] ANN Artificial Neural Network

[0071] PS Parameter Interface

[0072] MS Measurement Value Interface

[0073] LMS Learning Module Interface

[0074] CPU Central Processing Unit, Computing Unit

[0075] Se Sensor

Claims

1. A method for setting basic parameters of a component (K) in a test automation system (AA), wherein, The following method steps are performed before or during commissioning of the component (K) in the automation system (AA): - starting a test run (S2) of a component (K) in the automation system (AA) using the basic parameter setting, wherein the component (K) is an electronically controllable technical part or a field device; - measuring (S3) a measurement value data set during the commissioning; - accessing a machine learning module (ML) which is pre-trained in order to calculate (S4) a target parameter setting for the respective component (K) for a measured value data set, and wherein the basic parameter setting is compared with the calculated target parameter setting and, in the event of a deviation, a result message for adjusting the basic parameter setting is provided (S5), wherein a specific machine learning module (ML) is pre-trained for each individual component (K), wherein the machine learning module (ML) is trained to automatically recognize error patterns in the basic parameter setting, and wherein the following indicates that the component dimensioning is incorrect: (a) a very sudden stop in the final position of the pneumatic actuator by position measurement, (b) The controlled signal tends to oscillate, (c) excessive deviations in the set point value or actual value of the controlled signal, (d) unexpected force curves in extrusion applications, and / or (e) unexpectedly high friction in the case of incorrect mechanical design; and - Receive the provided result message to adjust the basic parameter settings (S6).

2. The method according to claim 1, wherein, The machine learning module (ML) is continuously retrained using continuously acquired measurement value data sets, the basic parameter settings and the calculated target parameter settings.

3. The method according to claim 1, wherein The machine learning module (ML) comprises a pre-trained neural network (ANN) which is trained using laboratory data and / or data from a test bench and / or data from at least one simulation model.

4. The method according to claim 1, wherein The machine learning module (ML) is designed to apply at least one pattern recognition algorithm to the measurement value data set.

5. The method according to claim 1, wherein The machine learning module comprises a component testing unit which is trained to test whether the corresponding component (K) is correctly inserted into the automation system (AA) in order to complete a predetermined automation task and, if the answer is no, to extend the result message with an exchange message.

6. The method according to claim 1, wherein The machine learning module (ML) is at least partially arranged locally on the component (K).

7. The method according to claim 1, wherein, The measurement data set includes signal characteristics that vary over time.

8. A parameter setting unit (P) for a component (K) used in an automation system (AA), the parameter setting unit being adapted to carry out the method according to claim 1 and comprising for this purpose: - Parameter interface (PS) for reading basic parameter settings; - at least one measured value interface (MS) for measuring and / or reading out measured value data sets; - an interface (LMS) of a machine learning module (ML), via which the result messages can be read to adjust the basic parameter settings; - A processor (CPU) that is used to control and operate the component (K) by using the basic parameter setting or the target parameter setting.

9. A component (K) for use in an automation system (AA) having a parameter setting unit (P) according to the previous claim.

10. The component (K) according to the previous claim, wherein, The component (K) includes a control unit, in particular a controller for an electrical or pneumatic system, and / or wherein the automation system (AA) includes electrical and / or pneumatic actuators.

Images