Using trained AI models in process for user management of field devices in automation technology

Abstract

The present invention comprises a computer-implemented process for training an AI model, comprising:-providing training data, where the training data comprises input data and output data, where the input data comprises identification and / or functional data of a plurality of field devices, and where the input data comprises identification and / or functional data of the plurality of field devices; the output data each including a result associated with the field device regarding whether login into the user management system is permitted; feeding the training data to the AI model; and-training the AI model using machine learning based on the training data to identify one or more relationships between the identification and / or functional data and the associated results, and using the trained AI model in a process for user management of field devices in automation technology.

Description

Technical Field

[0001] The invention relates to a computer-implemented process for training an AI model. Furthermore, the invention relates to the use of an AI model that has been trained with the aid of the process according to the invention in a user management process for field devices in automation technology. Furthermore, the invention comprises a computer-readable medium encoding instructions defining an AI model that has been trained using the process according to the invention. The invention also comprises a computer system comprising a computer-readable medium according to the invention. Background Art

[0002] Field devices used in industrial plants are already known from the prior art. Field devices are generally used in process automation engineering and manufacturing automation engineering. In principle, all process-oriented devices and all devices that provide or process process-related information are referred to as field devices. Therefore, field devices are used to detect and / or influence process variables. Measuring devices or sensors are used to detect process variables. These are used, for example, for pressure and temperature measurement, conductivity measurement, flow measurement, pH measurement, filling level measurement, etc., and detect corresponding process variables of pressure, temperature, conductivity, pH value, filling level, flow, etc. Actuators are used to influence process variables. These are, for example, pumps or valves that can influence the flow of fluid in a pipe or the filling level in a tank. In addition to the above-mentioned measuring devices and actuators, field devices are also understood to include remote I / O, radio adapters, or generally include devices arranged at the field level.

[0003] Many of these field devices are produced and sold by the Endress+Hauser Group.

[0004] In modern industrial plants, field devices are usually connected to a superordinate unit via a communication network such as a fieldbus (Profibus®, Foundation® Fieldbus, HART®, etc.). Typically, the superordinate unit is a control system (DCS) or a control unit such as an SPC (stored program control). The superordinate unit is used in particular for process control, process visualization and process monitoring as well as for the commissioning of field devices. The measured values ​​recorded by the field devices, in particular by sensors, are transmitted to one (or in some cases several) superordinate units via the corresponding bus system. In addition, data transmission from the superordinate unit to the field device via the bus system is also required, in particular for the configuration and parameterization of the field device and for controlling actuators.

[0005] The mobile control unit can also be used to operate field devices that have implemented the FDT frame application. For example, there are control units connected to a fieldbus network. However, the control unit can also communicate with the field device via a wireless communication connection (in particular based on the Bluetooth standard). The applicant produces and sells devices that allow the control unit to be coupled to the field device as a so-called Bluetooth gateway. The field device is connected to the Bluetooth gateway via a wired connection, in particular using the HART communication standard or a proprietary standard (for example, CDI, by means of which the applicant's field device can be addressed). Alternatively, the field devices themselves have their own Bluetooth interface.

[0006] If a mobile device such as a smartphone or a tablet computer is used as the control unit for wireless communication with the field device, application programs (so-called apps) are available, which make operating functions for the field device available to the mobile device.

[0007] In industrial environments, most installed field devices have no or only very basic protection against unauthorized access. In these field devices, it is usually possible to access all device parameters directly or, for example, after entering an unlock code. As a result of Germany's Federal Security Act, more and more field devices are appearing on the market with separate user accounts and role-based authorizations. For access via a user interface or machine interface, in a sense, "permanent" authorization is required, which is usually granted by a previous authentication. Authorization must be selected in such a way that the accessing user has (permanently) all the authorizations they need to perform their tasks. Standards such as IEC62443-4-2 require that "components" must have means of authenticating users and that there should be role-based authorization granting (authorization).

[0008] Due to the huge management effort of the login data of field devices, technologies for central user management are emerging, as has been common practice in the IT field for many years with respect to IT devices (e.g., printers, workstations, etc.). An example of such a concept is disclosed in DE 10 2018 102 608 A1, in which a transmission medium is provided, to which user data is transmitted from a user database, wherein access to the field device is granted after checking the user data.

[0009] There is also the idea of ​​limiting the access rights required by people to a minimum. DE102019131860 A1 discloses, for example, providing a digital order ticket transmitted from a server to a mobile device, which order ticket contains access rights and authorized tasks for a field device. When a connection is established with the field device, the order ticket is transmitted. If authorization is available, the field device can be used to process the tasks contained in the order ticket, such as the execution of parameterized actions or functional tests.

[0010] After delivery to the customer, field devices that can be managed with such an operating system must be connected to the unit to be managed (user management system in the form of a ticket server). This usually occurs in a key exchange between the field device and the management unit, which must be performed manually by the customer. The field device sends a join query to the unit to be managed. The user must then manually confirm the release. For a large number of field devices, this represents a personnel and cost expenditure that should not be underestimated, and it can also be prone to errors, as human error sources can inadvertently incorrectly classify field devices. There is also a risk of manipulation and / or cyberattacks, for example, if an attacker connects their own field device to the network and sends a join query to obtain information about users and rights. Summary of the invention

[0011] Based on this problem, the object of the invention is to connect a field device to a control system with user management in a simple manner.

[0012] This object is achieved by a computer-implemented process for training an AI model, wherein the process comprises the following steps:

[0013] - providing training data, wherein the training data comprises input data and output data, wherein the input data comprises identification and / or function data of a plurality of field devices, and wherein the output data each comprises a result associated with the field device as to whether a login into the user management system is permitted;

[0014] - Feeding training data to AI models; and

[0015] - Using machine learning to train an AI model based on the training data to identify one or more relationships between identification and / or functional data and associated outcomes.

[0016] The present invention assumes that, during the process of plant planning or design, there are certain structures or rules, so-called identification and / or functional data, which are available for each of the field devices. In addition, information is already available about which field devices are allowed to be added to the user management system and which field devices are not added to the user management system. These training data are used to teach or train the AI ​​model to recognize patterns or relationships in which field devices can be added and which field devices can not be added. This makes it possible to build a reliable decision-making system.

[0017] The field devices mentioned in connection with the invention have already been given as examples in the introductory part of the description.

[0018] A first variant of the process provides for a user to manually assign the results to each of the field devices, wherein the resulting training data after the assignment has been made is fed to the AI ​​model collectively as a training data set or individually.

[0019] According to a second variant of the process, the training data is fed to the AI ​​model in the form of a set as at least one training data set.

[0020] In both variants, the AI ​​model gradually identifies relationships between identification and / or functional data through this so-called supervised learning by learning to imitate the user's decisions and thereby training the classifier.

[0021] According to one embodiment, it is also possible to generate training data by dividing the identification and / or functional data into subsets, wherein the user assigns a result to each subset, wherein the training data sets each consist of a subset of the subsets of the identification and / or functional data and the results assigned to the subset. In a second variant embodiment of the process, it is provided that the division of the identification and / or functional data into subsets is performed with the aid of a further AI model. This is an AI model that is different from the actual AI model used in the process. The further AI model works here according to so-called unsupervised learning (also called "clustering"), in which a number of existing identification and / or functional data are checked for similarity and / or anomalies and grouped into subsets on this basis.

[0022] According to an advantageous embodiment of the process, the AI ​​model is based on a neural network. In particular, it is a perceptron or a "feedforward" network. Such a neural network has an input layer, an output layer and sufficient layers (so-called "hidden layers"), each layer having a large number of nodes or neurons between the input layer and the output layer. The input layer has one or more nodes or neurons for inputting identification and / or functional data. The output layer has nodes or neurons that output the result as "allowed" or "not allowed". Alternatively, it is also possible to provide an output layer that uses a numerical output for each possibility ("allowed" or "not allowed"), which represents the probability of this value. For example, if the probability of "allowed" is 55% and the probability of "not allowed" is 45%, a closer look or inspection is needed, while if the probability is 99% or 1%, the result of the model has more confidence.

[0023] Neural networks are also used in further AI models.

[0024] According to one embodiment of the process, the identification and / or functional data includes one or more of the following data categories:

[0025] - Name of the measuring point, TAG

[0026] - Network address of the field device;

[0027] - Name of manufacturer;

[0028] - Type of device;

[0029] - Serial number;

[0030] - Device name.

[0031] It is obvious to a person skilled in the art that further identification and / or functional data relating to field devices, measuring points or networks of the system can also be used for the training data in order to be able to recognize connections to user decisions. It is important to ensure that the training data contains a sufficiently large amount of such data (preferably at least 1,000 different data items per category, particularly preferably at least 10,000 different data items per category).

[0032] Furthermore, the object is achieved by using an AI model which has been trained by means of a process according to the invention in a process for user management of field devices in automation technology, wherein the process comprises the following steps:

[0033] - making the subscriber management system known in the automation system by sending an invitation telegram to at least one field device used in the system;

[0034] - in response to the invitation telegram, sending a registration query of the field device to the user management system, wherein the registration query contains identification and / or functional data of the field device;

[0035] - checking identification and / or functional data by the user management system using an AI model to determine whether to allow the field device to log into the user management system; and

[0036] - In case the check produces a successful result, the field device is registered with the user management system.

[0037] This allows reliable decisions to be made as to whether a new field device can be added to the user management system. The process can be performed quickly and is highly accurate.

[0038] One embodiment of use provides that, before the step of checking the identification and / or functional data by the user management system, the user first manually checks with the aid of the AI ​​model whether the field device is allowed to log into the user management system, wherein the field device is only logged into the user management system if the user's check and the AI ​​model's check each produce a successful result. Thus, the AI ​​model is used to check or verify the user's manual decision about the registration and to provide a second opinion. This can increase the safety and reliability of manual decisions made by the user.

[0039] If the AI ​​model is trustworthy, field devices can also automatically register with the user management system after their decision without human assistance.

[0040] According to one embodiment of the use, it is provided that at least one field device and the user management system are communicatively connected via a local network or via the Internet. The communication connection can be wired or wireless. In a wired variant, an Ethernet-based network can be used as the local network. Alternatively, an automation fieldbus can be used, in particular one based on the standards Profibus PA / DP, Foundation fieldbus, etc. In the case of a wireless design, for example, a network based on one of the standards WiFi, WirelessHART, ZigBee or Bluetooth can be used.

[0041] The object is also achieved by a computer-readable medium encoding instructions defining an AI model that has been trained by means of the process according to the invention.

[0042] The object is also achieved by a computer system comprising a computer-readable medium according to the present invention. The computer system executes an AI model and / or trains an AI model, and is thereby able to perform a training process and the use of the trained AI model in a decision-making process. For example, the computer system can be a user management system that is locally arranged near the system and can communicate with the system and field devices and / or control units located therein, for example via a LAN connection. Alternatively, the user management system is cloud-based and can be contacted via the Internet. Alternatively, the computer system that trains and / or executes the AI ​​model is not a user management system, but a separate component. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The present invention is explained in more detail with reference to the following drawings. In the drawings:

[0044] Figure 1 A first exemplary embodiment of a process according to the invention is shown, wherein the AI ​​model is taught using training data; and

[0045] Figure 2 An exemplary embodiment for generating training data is shown. DETAILED DESCRIPTION

[0046] If the field device FG is to be added to a user management system (e.g., a ticket server), the user management system sends an invitation telegram via all communication connections available to the user management system. Those field devices FG that are compatible with the user management system send registration queries AA1, ..., AAn back to the user management system as response telegrams.

[0047] The user BN (eg an administrator of a subscriber management system) must now check the individual registration queries AA1 , . . . , AAn and decide individually whether the corresponding field device FG is to be added to the subscriber management system.

[0048] A perfect and safe selection of the field device FG could be based on an exact comparison of keys, in particular ED25519 keys. However, this is impractical because, on the one hand, these keys are not known in advance and, on the other hand, they are not manageable for humans due to the large number of characters. Therefore, the user BN tries to use other information from the registration query of the field device FG.

[0049] For a specific test of the field device FG, the user BN preferably consults a parameter from the "PostalAddress" group, which is part of the registration queries AA1, ..., AAn. The following is an extraction of "PostalAddress" from the registration queries AA1 to AAn:

[0050]

[0051] This "PostalAddress" contains one or more identification and / or functional data items, which are used to identify the field device itself or a function in a specific system, or to describe the type of communication connection of the field device FG in the system network. (The identification and / or functional data contained in this example are, for example, the network address of the field device FG (="fieldbus-address" or "ip-address", the device name (="device-id", etc.), etc.). This list only represents an excerpt of the identification and / or functional data contained in the registration queries AA1, ..., AAn. It can be provided that the registration queries AA1, ..., AAn contain alternative and / or additional data fields to the "PostalAddress", which also contain identification and / or functional data.

[0052] The user now starts to decide for each field device FG which field devices FG are allowed to log into the user management system and which are denied. This means that the user BN marks the data. Specifically, he marks the data, in which he assigns the result "allowed" or "not allowed" regarding the login to the user management system. Thereafter, the marked data will be referred to as "training data". Therefore, the training data contains the identification and / or functional data of the field device FG as input data ED, and the result or label as output data AD. After creation, these training data are gradually fed into the AI ​​model KI1 and used to train the AI ​​model KI1. The training data can also be collected as training data sets TD1, TD2 and fed to the AI ​​model KI1.

[0053] The AI ​​model KI1 is based in particular on a neural network. In particular, it is a perceptron or "feedforward" network. Such a neural network has an input layer, an output layer and sufficient layers (so-called "hidden layers"), each layer having a large number of nodes or neurons between the input layer and the output layer. The input layer has one or more nodes or neurons for inputting identification and / or function data. The output layer has nodes or neurons that output the result as "allowed" or "not allowed".

[0054] The AI ​​model KI1 can be trained using training data. In a process called machine learning, the AI ​​model KI1 learns decision-related patterns. The AI ​​model KI1 learns the relationship between input data ED (identification and / or functional data) and associated output data AD (evaluation results of the user BN).

[0055] Figure 2 Another example of how to create a training data set TDS1, ..., TDSn that is fed to an AI model KI1 for training is described. In this example, a large number of registration queries AA1, ..., AAn from a large number of field devices FG have arrived at the user management system after the system has been put into operation. Another AI model KI2 that works on cluster detection / pattern recognition analyzes these registration queries AA1, ..., AAn and starts clustering them. This involves dividing into several (abstract) clusters or groups. For example, certain items of identification and / or functional data of all registration queries AA1, ..., AAn in a group are similar.

[0056] Then, user BN actively labels each of these groups with a result of "allow" or "not allow". Thus, a training dataset TDS1, ..., TDSn is created from each group.

[0057] In the further process, all field devices FG whose registration queries AA1, ..., AAn are in the group with the result "allowed" can be admitted to the user management system. In addition, separate training data sets TDS1, ..., TDSn are passed to the AI ​​model KI1 for training.

[0058] The trained AI model KI1 can then apply what it has learned and support or replace the user in making decisions regarding new field devices. To this end, the AI ​​model checks the registration queries AA1, ..., AAn of the new field device FG.

[0059] "Supporting" user BN means that the AI ​​model creates a second opinion in addition to the manual decision of user BN. Therefore, AI model KI1 is able to confirm or deny the manual decision of user BN.

[0060] "Replacing" the user BN means that the AI ​​model has the sole decision-making ability to allow or deny the field device FG to log into the user management system. Therefore, the registration process can also be performed automatically.

[0061] The invention promises the advantage of being able to save time when registering a field device FG with a user management system, while at the same time increasing the reliability of the decision quality compared to a human user BN as decision authority.

[0062] Reference numerals list

[0063] AA1, ..., AAn query response

[0064] AD output data

[0065] BN User

[0066] ED Input Data

[0067] FG Field Devices

[0068] KI1 AI Model

[0069] KI2 Other AI models

[0070] TDS1, …, TDSn training data sets

Claims

1. A computer-implemented process for training an AI model (KI1), comprising: - providing training data, wherein the training data comprises input data (ED) and output data (AD), wherein the input data (ED) comprises identification and / or function data of a plurality of field devices (FG), and wherein the output data (AD) each comprises a result associated with the field device (FG) as to whether a login into the user management system is permitted; - feeding the training data to the AI ​​model (KI1); and - Training the AI ​​model (KI1) using machine learning based on the training data to identify one or more relationships between the identification and / or functional data and associated outcomes.

2. The process of claim 1, wherein: A user (BN) manually assigns results to each of the field devices (FG), wherein the resulting training data is fed individually to the AI ​​model (KI1) after the assignment has taken place.

3. The process of claim 1, wherein: The training data, grouped into at least one training data set (TDS1, ..., TDSn), are fed to the AI ​​model (KI1).

4. The process of claim 3, wherein: The identification and / or functional data are divided into subsets, wherein a user (BN) assigns results to each subset, wherein the training data set or multiple training data sets (TDS1, ..., TDSn) each consist of a subset of the subsets of the identification and / or functional data and the results assigned to the subset.

5. The process of claim 4, wherein: The division of the identification and / or functional data into the subsets is performed with the aid of a further AI model (KI2).

6. The process according to one of the preceding claims, wherein: The AI ​​model (KI1) is based on a neural network.

7. The process according to one of the preceding claims, wherein: The identification and / or functional data includes one or more of the following data categories: - Name of the measuring point, TAG - the network address of the field device (FG); - Name of manufacturer; - Type of device; - Serial number; - Device name.

8. Use of an AI model (KI1) which has been trained by means of a process according to one of claims 1 to 7 in a process for user management of field devices (FG) in automation technology, wherein: The process includes the following steps: - making the subscriber management system known in the automation system by sending an invitation telegram to at least one field device (FG) used in the system; - in response to the invitation telegram, sending a registration query (AA1, ..., AAn) of the field device (FG) to the subscriber management system, wherein the registration query (AA1, ..., AAn) contains identification and / or functional data of the field device (FG); - checking the identification and / or functional data by the user management system using the AI ​​model (KI1) to determine whether the field device (FG) is allowed to log into the user management system; and - In case said check yields a successful result, registering said field device (FG) with said user management system.

9. The use according to claim 8, wherein: Prior to the step of checking the identification and / or functional data by the user management system, the user (BN) first manually checks with the aid of the AI ​​model (KI1) whether the field device (FG) is allowed to log into the user management system, wherein the field device (FG) is only logged into the user management system if the check by the user (BN) and the check by the AI ​​model (KI1) each produce a successful result.

10. The use according to claim 8 or 9, wherein: The at least one field device (FG) and the user management system are communicatively connected via a local network or via the Internet.

11. The use according to claim 10, wherein: The communication connection is wired or wireless.

12. A computer-readable medium encoding instructions defining an AI model (KI1) trained by a process according to any one of claims 1 to 7.

13. A computer system comprising the computer readable medium according to claim 12.

Images