Method and system for supported control of a machine line

Abstract

An embodiment of the invention relates to a method and a system for controlling a machine line, in particular in a machine line for filling and packaging food and / or beverages. The machine line comprises a plurality of machines. The embodiment comprises implementing a large language model (LLM) on a server, and determining a control command context and a specific context for the LLM. The control command context comprises a set of control commands available for the machine line and their syntax. The specific context comprises a restriction of an output of the LLM. After inputting status data of the machine line into the LLM, a user request to output a control command for the machine line can be received via a user interface. The LLM processes the user request on the basis of the determined control command context and the specific context and on the basis of the status data, and creates a response. The response comprises a control command which is generated by the LLM and can be executed by the machine line.

Description

[0001] Method and system for assisted control of a machine line

[0002] The invention relates to a method and a system for the supported control of a machine line, in particular a machine line for filling and packaging food and / or beverages, and a machine line.

[0003] Today's filling and packaging systems in the liquid food industry are highly optimized and process up to 120,000 units per hour. A typical filling and packaging system typically comprises a multitude of different machines and modules connected by conveyor belts. The units, such as bottles, cans, pallets, containers, or similar, are transported from one machine to the next, passing through the individual machines in the line in a predetermined sequence.

[0004] Numerous evaluations have shown that once the factory acceptance test is completed on-site at the customer's site and the technicians leave the construction site after assembling the machine line (e.g., the beverage bottling line), the customer's overall equipment effectiveness (OEE) often plummets. The reasons for this are varied. They can be due to a lack of user knowledge, incorrect machine operation, inefficient troubleshooting, and many other causes.

[0005] For these reasons, digitalization solutions are often provided that summarize data from a beverage bottling plant at a higher level and provide insight into production. Customers can then view key production metrics such as "number of malfunctions per machine," "number of containers produced," "number of containers rejected," "OEE," etc., analyze them in more detail, and derive measures to prevent problems / malfunctions from occurring in the future or to resolve them quickly if they do occur.

[0006] In the age of digitalization, the automation of complex machinery has become increasingly important. Traditional control systems rely on preprogrammed algorithms and fixed schedules to operate machines. Although these systems are sophisticated and efficient, they often lack the self-learning and adaptability required to cope with unexpected situations or changes in the production environment. In this context, artificial intelligence (AI) has been increasingly used in recent times. AI-based systems are capable of learning from data, recognizing patterns, and making decisions based on them. They can continuously learn from their performance and improve over time, making them ideal for controlling complex machinery. Nevertheless, existing AI-based control systems are still not perfect.

[0007] The training of AI for machine control relies on the availability and quality of data. AI models typically require a large amount of high-quality training data to make accurate predictions and effective decisions. This data is often obtained from the operating history of the machine and from experiments under various operating conditions.

[0008] However, highly specialized machinery presents significant challenges. For example, the amount of available training data can be limited. This is because specialized machinery is often designed for specific applications and therefore unable to experience a wide range of operating conditions. Furthermore, such machines are often expensive and difficult to operate, making it difficult to conduct data collection experiments.

[0009] Furthermore, the quality of the available data can often be poor. Highly specialized machinery, such as that found in modern beverage bottling plants, is typically complex and has many variable parameters, making it difficult to collect accurate and complete data.

[0010] These limitations make training AI models for the control of highly specialized machinery a significant challenge. Traditional AI models, which rely on learning from large amounts of training data, often cannot function effectively in such environments. They can struggle to make accurate predictions and effective control decisions, leading to suboptimal performance and inefficient operation.

[0011] It is therefore necessary to provide a more efficient way to achieve computerized control of complex machine lines.

[0012] There is therefore a need for improved systems and methods for computer-aided control of a machine line. This object is achieved according to the invention by a method according to claim 1 and a system according to claim 7. Embodiments and further developments are covered in the subclaims.

[0013] One embodiment of the invention relates to a method for controlling a machine line, in particular in a machine line for filling and packaging food and / or beverages. The machine line comprises a plurality of machines. The method comprises implementing a large language model (LLM) on a server and defining a control command context and a specific context for the LLM.

[0014] A server can be configured as a server, an industrial PC, or an edge device, or comprise both. The specific configuration depends, for example, on the requirements of the LLM model used and / or the type of communication. For example, it is possible to host the LLM model externally, i.e., with a service provider such as the operator of a server farm. Communication with the machine line can then take place via an API. However, the LLM model can also be hosted internally by the operator of the machine line.

[0015] The control command context includes a set of control commands available for the machine line and their syntax. The specific context defines a restriction on the output of the LLM. After inputting state data of the machine line into the LLM, a user request to issue a control command for the machine line can be received via a user interface. The LLM processes the user request based on the specified control command context, the specific context, and the state data and creates a response. The response includes a control command generated by the LLM that can be executed by the machine line.

[0016] One embodiment of the invention relates to a system comprising an LLM, a server and a user interface.

[0017] Exemplary aspects of the invention are illustrated in the drawings. They show:

[0018] Figure 1 : a diagram showing an overview of the essential elements and the basic structure of the invention;

[0019] Figure 2: a diagram showing a gradual adaptation of the large language

[0020] Model for use in a machine line; Figure 3: an exemplary flow chart for a method for

[0021] Implementing a large language model and its operation in a machine line;

[0022] Figure 4: an exemplary system configuration for PET containers and

[0023] adhesive packaging;

[0024] Figure 5: an exemplary plant configuration for PET containers and

[0025] shrink packer;

[0026] Figure 6: an exemplary system configuration for cans or glass bottles; and

[0027] Figure 7: an example system configuration for cans.

[0028] Figure 1 is a diagram showing an overview of the essential elements and basic structure of the invention. According to one embodiment, these essential elements may be a machine line 100, a cloud or computer server 110, and a user interface 120 via which a user 130 can interact with the server 110 and, if appropriate, also with the machine line 100.

[0029] The machine line 100 may, for example, be a machine line for filling and packaging food and / or beverages, but is not specifically limited to such a line. Various exemplary embodiments of such a machine line 100 are shown and explained in Figures 4 to 7.

[0030] The computer server 110 is not necessarily a single server, but may consist of one or more computing devices that are communicatively connected directly or indirectly. As described in more detail below, the server 110 may include a large language model (LLM) and provide an interface between the LLM and a user interface 120. The server 110 may be communicatively connected directly or indirectly to the machine line 100. According to embodiments, an internet connection exists between the server 110 and the machine line 100, through which they are connected to each other, or there may also be a direct communication connection. That is, the LLM on the server 110 and the machine line 100 can communicate directly with each other.The communication between server 110, the LLM and the machine line 100 can be monitored or controlled by a user 130 via the user interface 120.

[0031] The user interface 120 allows the user 130 to interact with the server 110 or at least with certain services on the server 110. The user interface can take any form suitable for a user to interact with a computer by means of inputs. The user interface is also referred to as a "human-machine interface" (HMI) and, in certain circumstances, allows the user 130 to observe the system status and intervene in the process beyond operating the machine line 100.

[0032] The user interface may be a simple computer, a mobile phone, a tablet, a special device of the machine line 100, a voice assistant, a computer application, a computer terminal or another computer-based device.

[0033] As already mentioned, a so-called large language model (LLM) can be implemented on server 110. Such large language models, or language models, are a type of artificial intelligence (AI) trained to understand and generate human language. They learn by analyzing large amounts of text data and recognizing patterns within it. This involves not only individual words and sentences, but also the context in which they are used. Special language models called "Transformers" can recognize more complex contexts and also learn to generate answers to questions, translate text, or even create graphics such as plots and diagrams. The well-known architecture called "Transformer," based on a method called "Attention," helps the language model understand the context of a word in relation to all other words in a text passage.

[0034] Such a neural transformer consists of a number of layers, each containing a set of neurons. Each layer and each neuron is adapted during the training process to better respond to the training data. The "depth" of the model (the number of layers) and the "width" (the number of neurons in each layer) determine the complexity of the patterns the model can recognize.

[0035] The training process requires a large amount of computing power and time, but once the model is trained, it can be used to generate or interpret text based on the patterns it has learned. Such LLMs are provided pre-trained by various vendors, allowing end users to use the "finished" LLMs.

[0036] Embodiments of the invention relate to the integration of such LLMs into a user platform to provide computer-based control or at least computer-assisted control of the machine line 100. According to embodiments of the invention, the LLM is adapted in a special way to provide the user 130 with requested data from the machine line 100. The requested data can be processed information about the machine line 100, but can also be control commands executable by the machine line 100. When adapting the LLM, security aspects and data transparency are taken into account to ensure that no incorrect information is provided to the user and that the control commands do not initiate unwanted processes in the machine line that could potentially lead to damage to the machine line 100.

[0037] This concept of adapting the LLM is shown in more detail in Figure 2. The architecture in Figure 2 is essentially the same as in Figure 1 and includes the machine line 100, the server 110, and the user interface 120. Additionally, Figure 2 shows the adaptation of an LLM 200, which is implemented as an example on the server 110, and the output of control commands from the LLM 200 to the machine line 100.

[0038] The LLM 200 can be a general, fully trained LLM, which in its initial state may not have any advanced knowledge of the machine line 100 or the data produced by the machine line. Therefore, it is first necessary to give the LLM a context.

[0039] A fundamental idea in applying the LLM 200 to the control of the machine line 100 is to describe the control rules and principles of the plant in a formalized language and to teach this description to the LLM. Instead of requiring specific training data for the machine line 100 representing the various operating conditions of the machine plant, the LLM 200 learns how to control the machine line 100 from these formalized rules, i.e., from control commands and their syntax, and from the operating history (e.g., from state data and / or the history of the state data and the associated control parameters associated with the corresponding states).

[0040] This has several advantages. The LLM 200 does not require specific training data for the machine line 100, which circumvents the problem of limited data availability in highly specialized machine systems. Furthermore, because the LLM 200 is based on text processing, it can be easily adapted and updated by adding new rules, parameters, syntax, or principles. Furthermore, the LLM 200 is capable of recognizing and learning complex relationships and patterns within the aforementioned rules and relationships, enabling it to make effective control decisions even in complex and dynamic operating environments. Furthermore, the LLM 200 has the advantage of being transparent and comprehensible. The decisions it makes are based on clear and explicit rules, a learned syntax that can be easily understood and verified.This is especially important in safety-critical applications where it is essential to understand and monitor the behavior of control systems.

[0041] Thus, the use of LLMs described herein to control or assist in the control of the machine line 100 offers an innovative and promising solution that overcomes the challenges of traditional AI models and paves the way for more efficient, reliable, intuitive, and adaptable control systems.

[0042] However, to implement an LLM 200, it must first be prepared for the specific use for (supporting) control.

[0043] The "general LLM" 200 thereby becomes an "LLM with context" 200', as shown in Figure 2. The context may consist of a specific context and a control command context, as further described below.

[0044] The specific context can be specific to a very specific system configuration of machine line 100 or specific to specific user groups. The specific context defines a restriction on the LLM output. Such restrictions can, for example, concern a form, a structure, a used information source, or a level of detail of the LLM response. Furthermore, the specific context can also include special instructions or guidance for calculating certain key figures or information.

[0045] To precisely define the specific context for the LLM, appropriate, specific textual instructions can be defined. These can also be user-specific (e.g., which content can be viewed by a customer, a plant manager, or a technician). The form and structure of the desired LLM responses can also be defined here.

[0046] In order to make it possible to find the cause of a particularly low value, for example, in the case of calculated key figures, the LLM can also be provided with facts about the calculation of the key figures and causal relationships.

[0047] Since the fully trained LLM "understands" and accepts human language as input, the specific context can also be formulated as such a text input in one embodiment. A simplified, exemplary specific context could be as follows:

[0048] "You are an intelligent assistant named Chat2Act that helps users with their questions about the data produced by their machine lines and can also issue control commands for the machine line. Answer ONLY with the facts listed in the list of sources provided to you. Only create control commands that can be justified by these facts. If this information is insufficient, say you don't know. Do not generate answers or control commands that do not relate to the provided sources or control command context. If a clarifying question to the user would be helpful, ask it."

[0049] This specific context is merely exemplary and may be even more detailed and restrictive in the actual implementation. The specific context is supplied to the LLM 200', for example, by a manufacturer of the machine line 100. The specific context may be entered into the LLM 200' in such a way that it is not possible for a user 130 who later interacts with the LLM to change the specific context in such a way that restrictions established by the specific context can be removed again.

[0050] As mentioned above, the LLM 200 is also given a control command context, which describes the rules for issuing the control commands in more detail. The control command context can include a set of control commands available for the machine line and their syntax. For example, the control command context can specify one or more parameter value ranges for the set of control commands within which parameters may be changed by the control command. The control command context can also prohibit certain parameter ranges or parameter changes, for example, for safety reasons, as these could result in damage to the machine line 100.

[0051] However, the context does not necessarily have to be divided into a specific context and a control command context. Both contexts can also be combined into a single context.

[0052] However, the "LLM with context" 200 cannot yet be used to obtain specific control commands for the machine line 100, since in a further step, the data or status data of the machine line 100 must first be entered into the LLM (e.g., as sources). The data 230 can, for example, be loaded directly from the machine line 100 into the cloud 110 or can also be transferred to the cloud 110 in a controlled manner by a third party. This data 230 forms the database for deciding how to respond to user requests and for generating control commands for the machine line 100. In addition to the control command, which can be sent directly to the machine line 100, the response can also include directly collected or calculated key figures, sensor data, and / or machine data from the machine line 100 or individual machines of the machine line 100, as described later.For this purpose, an existing database can be used, which also serves as a source for existing services.

[0053] According to embodiments, it may be necessary to extract suitable data from the database 231, e.g., data for a particular production line for a particular time period or multiple time periods.

[0054] According to embodiments, it may further be necessary to suitably transform the data 231 extracted from the database for use in an LLM so that it can be processed by the latter. This processing step is shown within the data box 230, which shows raw data 231 that is converted into the processed data 232 by extracting and processing suitable contents of the data 231. This may include converting values ​​into suitable units (e.g., timestamps, time differences, percentages), calculating ratios, and restructuring larger, multidimensional data sets. The processing may also include adding data context to the data, such as labels about the content and meaning of the data, so that the LLM can correctly interpret the data.

[0055] This data preprocessing, as shown in data box 230, can be performed in regular intervals and can be regularly precalculated and persisted for specific periods of time to minimize the latency of the LLM's processing of the user request. In other words, processing data over longer periods of time (weeks, months, quarters) can take a long time. Therefore, in these cases, the data should already be precalculated and persisted to minimize the waiting time for the user 130.

[0056] After completing the input of the status data into the LLM with context, an "LLM with context and with status data" 200 is created, which is ready for use in the cloud 110. The LLM 200 can be called up by a user 130 via the user interface 120 and can be used as a digital assistant for the customized output of data and information from the machine line and, according to the invention, can also be used for the direct control of the machine line 100.

[0057] The user 130 can now interact with the LLM 220 and send requests to the LLM via the user interface 120 to obtain requested data from the LLM 200. This data can be information about the productivity of the machine line 100, but can also be control commands to be issued by the LLM 200. The control commands can be communicated directly to the machine line 100 and executed there (e.g., after user approval) by the machine line.

[0058] The structured output / response of the LLM 200 can be used to provide data to third-party systems (which do not understand natural language). For example, this can be used to create instructions for users in the machine line 100 or to store structured information for documentation purposes. Furthermore, this structured data can be used to control the machine line 100 using suitable parameters.

[0059] This is shown in Figure 2 by the data packet 240, which is sent from the LLM 200 directly to the machine line 100 and includes one or more instructions or control commands for setting the machine line 100. The instructions or control commands are in a machine-readable file format. The machine-readable file format can be, for example, JSON or XML.

[0060] The usability of the user platform of the machine line 100 is thus improved because the user 130 is supported by the LLM in a question-and-answer style and is able to navigate more quickly to the desired information or to receive a desired control command more quickly, with which the machine line 100 is controlled according to the user's wishes.

[0061] According to a further aspect of the invention, the response of the LLM 200 may also include instructions to a user, e.g., in addition to or instead of the control commands sent directly to the machine line 100.

[0062] When issuing instructions to the user, multiple options can also be provided. Especially with complex control systems such as reacting to events across an entire machine line, there isn't just one solution to resolve the problem. Often, setting one value requires other, further changes to the process.

[0063] In this respect, a model as versatile as the LLM 200 makes it possible to offer alternative courses of action. For example, in response to an analyzed impending malfunction, it is possible to either propose maintenance as quickly as possible while continuing operations at the same (maximum) speed. It is also possible to postpone maintenance and reduce the output of the entire line to avoid damaging the part requiring maintenance.

[0064] These decision options can be made available to the user through the LLM 200. However, since the LLM 200 can access all data, the impacts of the respective alternatives on the line (e.g., OEE, postponement of subsequent production runs, etc.) can be specified in addition to the course of action. This makes it easier for the user to select the alternatives.

[0065] Figure 3 shows an exemplary flowchart for a method for implementing an LLM and its operation in a machine line 100.

[0066] The method 300 begins with step S302, in which a large language model is implemented on the server 110. The implementation may include the implementation of a specific general language model that is optimized for a specific language (English, German, French, Spanish, Chinese, etc.) or that supports multiple languages.

[0067] In step S304, a context for the LLM is defined. The context may consist of a specific context and a control command context. The specific context may specify a restriction on an output of the LLM. The specified restriction may restrict at least one of the following: a form, a structure, a utilized information source, and a level of detail of the LLM output.

[0068] The control command context includes a set of control commands available for the machine line and their syntax, which is taken into account when the LLM creates control commands.

[0069] Both the specific and the control command context can be defined and entered into the model by a manufacturer of the machine line 100, an operator of the online service, or another entity. In step S306, status data is entered into the LLM for the machine line 100. The status data can include directly collected or calculated key figures, sensor data, and / or machine data. The status data can be from one or more specific time periods. In an optional substep of S306, this status data can be preprocessed.Preprocessing the state data for the machine line 100 may include at least one of the following operations: (i) converting one or more values ​​in the state data into one or more suitable units that can be processed by the LLM, (ii) calculating ratios, (iii) restructuring multidimensional data sets, and / or (iv) adding data context to the data.

[0070] In the following step S308, a user request to issue a control command for the machine line is received via the user interface 120. For example, the user request from user 130 may be a request that the LLM issue a control command to the user or the machine line 100 that resolves a previously identified problem in the currently running operation. This request is only an example.

[0071] The user request is entered into the LLM and processed by the LLM in step S310 based on the specified specific context and based on the state data.

[0072] Finally, in step S312, a response to the user request is created. The response may include a control command generated by the LLM that is executable by the machine line. The provision may further include forwarding it to the user 130 from whom the user request originated. The provision of the requested information by the LLM may further include providing information about one or more sources used by the LLM to determine the requested information. The information about the one or more sources may appear textually and be selectable by the user 130 and linked to the respective sources. For example, the user may click on the source and, in addition to the LLM's response, be taken to the database that serves as the basis for the LLM's answer to the question.

[0073] In the final step S314, the generated control command can be output to the machine line 100. According to one embodiment, however, the control command is not automatically executed by the machine line 100, but only after receiving confirmation from a user that the control command should be executed by the machine line. The following Figures 4 to 7 describe various exemplary system configurations for various bottle filling systems in which the invention, or at least parts and aspects of the invention, can be implemented. The description of Figures 4 to 7 is intended only to provide a general overview of machines for which status data can be collected, on the basis of which the LLM can process user requests.

[0074] Figure 4 shows an exemplary system configuration 1000 for PET bottles or PET containers and adhesive packs. As shown in Figure 4, the system configuration 1000 comprises various modules that form a line, at the end of which the finished PET containers are dispensed in the form of a pack on pallets. Some of the modules and machines may be optional, and the invention is not limited to the exact form and arrangement of the system configurations.

[0075] The system configuration 1000 comprises an oven 1002 for preforms, a preform sorter with a feeding machine 1004, and a blow molding machine 1008. The modules 1002, 1004, and 1008 generally form a stretch blow molding machine in which PET containers are produced and formed from a starting material. The produced PET containers are forwarded to a filler 1010, where the bottles are filled. The filler can optionally include a rinser. Various particles such as dust, cardboard, or remnants of wooden pallets can settle in the preforms during storage or transport. These can be removed with the rinser. A closer can be arranged at the end of the filler, by means of which the PET containers are closed after filling.

[0076] Optionally, the system configuration 1000 can include a rotating device downstream of the filler 1010, which is used for hot filling of the PET containers. Via one or more conveyor belts 1016, which can also include a buffer 1018 for intermediate loading of filled containers, the filled PET containers are conveyed to a separator 1020 and then to a drying device 1024, in which the PET containers are dried.

[0077] After drying, the PET containers are conveyed to a labeling machine 1026. The labeling machine 1026 can be designed for various labeling techniques, such as labeling using hot melt, cold melt, self-adhesive labels, or sleeves. After the PET containers have been printed or labeled, they are conveyed through a second drying device 1028, a line distributor 1030, conveyor belts 1032, an adhesive pack production line 1034, and a curing section to a handle applicator. In the adhesive pack production line 1034, the PET containers are grouped into specific group sizes and packaged into a pack, such as a "six-pack." In the handle applicator, a carrying handle is attached to the pack, which allows for comfortable carrying of the pack.The finished containers are then arranged accordingly by a robot 1042 for layer production and packed on pallets by a palletizer 1044.

[0078] In system configuration 1000, so-called format trolleys or format racks can be arranged on various modules and machines to provide quickly interchangeable format sets for short changeover times and automatic tool changes. Examples of format trolleys are format trolley 1006 for blow molding machine 1008, format trolley 1012 for filler 1010, format trolley 1022 for labeling machine 1026, format trolley 1038 for adhesive pack production 1034, and format trolley 1046 for palletizer 1044.

[0079] Figure 5 shows another example system configuration 1100 for PET containers and shrink packers. System 1100 from Figure 10 includes many of the modules and machines from system configuration 1000 from Figure 4, but there are some differences. Therefore, the description of the modules already described in connection with Figure 4 is omitted for Figure 10.

[0080] A key difference between the two exemplary system configurations 1000 and 1100 is that the labeling machine 1126 with the labeling modules 1127 can be installed downstream of the blow molding machine 1008 and upstream of the filler 1008. For this purpose, the system configuration 1100 can comprise six transport lanes 1150 into which the PET containers can be pushed. After the PET containers have pushed into one of the six lanes 1150, they are conveyed into the film wrapping module 1152 and then into the shrink tunnel 1154.

[0081] Figure 6 shows an example system configuration 1200 for cans or glass bottles. The example system configuration 1200 from Figure 6 again has some similarities to the system configurations 1000 and 1100 from Figures 4 and 5, and the description of the system configuration is therefore limited to the differences between the system configurations.

[0082] As shown in Figure 6, the exemplary system configuration can include two separate feeders. A first feeder, on the left in Figure 6, shows a branch for cans or, optionally, a partial branch for new, reusable bottles. The containers, i.e., cans or new bottles, are fed into the machine by a depalletizer 1302, where they are guided via conveyor belts to the filler 1010. A second feeder, on the right in Figure 6, shows a partial branch for reusable bottles, which are fed into the system by a reusable sorting system (not shown).

[0083] In the case that the already used reusable bottles are introduced into the system 1200 via the sub-branch for reusable bottles, the reusable bottles first pass through the cleaning machine or washing machine 1304. Another possible difference in the exemplary system configuration 1200 is the transfer packer 1306 after the labeling machine 1026. The transfer packer can sort the bottles or cans into a carton clip application or into crates, or both.

[0084] Figure 7 shows an exemplary system configuration 1300 for cans, in which the elements already described in the other system configurations are no longer described. The cans in system configuration 1300 are fed into the depalletizer 1302 from a magazine 1402 containing cans. After passing through the filler and being filled, the cans are closed by means of a closure magazine 1404 and transported further along the system 1400 via the conveyor belts, as described above.

[0085] The optional Pasteurizer 1408 can be bypassed via the Bypass 1412 if not required. Freshly filled products can be pasteurized in the Pasteurizer 1408 for preservation.

[0086] In contrast to plant configurations 1000, 1100, and 1200, the exemplary plant configuration 1300 shows various tanks for corresponding consumables, such as tanks 1410 with rinsing liquid and / or the filling product and tanks 1406 with belt lubricant. These tanks can also be included in the exemplary plant configurations described above. For example, the chemical products 106 that are fed from the mixer 110 to the machines can be stored in tanks 1406 and 1410.

Claims

CLAIMS 1. A method for controlling a machine line (100), in particular in a machine line for filling and packaging food and / or beverages, wherein the machine line comprises a plurality of machines, and wherein the method comprises: Implementing (S302) a large language model, LLM, on a server (110); Defining (S304) a control command context and a specific context, wherein: the control command context comprises a set of control commands available to the machine line and their syntax, the specific context defines a restriction on an output of the LLM, and the defined restriction restricts at least one of an output format, a structure, and a used information source for a response of the LLM; Entering (S306) machine line status data into the LLM; Receiving (S308), via a user interface (120), a user request regarding the machine line; Processing (S310) the user request by the LLM based on the specified control command context, the specific context and based on the state data; Creating (S312) a response to the user request based on the processing, wherein the response comprises a control command generated by the LLM that is executable by the machine line; and Output (S314) of the generated control command by the LLM.

2. The method of claim 1, further comprising: Receiving confirmation from a user that the control command should be executed by the machine line.

3. The method according to claim 1 or 2, wherein the control command context further specifies one or more parameter value ranges for the set of control commands within which parameters may be changed by the control command.

4. The method according to any one of claims 1 to 3, wherein the control commands are output in a machine-readable file format, and wherein the machine-readable file format is JSON or XML.

5. The method according to any one of claims 1 to 4, wherein the status data comprises at least directly collected or calculated key figures, sensor data and / or machine data, and wherein the status data are from one or more specific time periods.

6. The method according to any one of claims 1 to 5, further comprising: Preprocessing the condition data for the machine line, wherein the preprocessing of the condition data comprises at least one of: Converting one or more values in the state data into one or more suitable units that can be processed by the LLM, Calculating ratios, Restructuring multidimensional data sets, Aggregating data, and Adding data context to the data.

7. System for controlling a machine line, in particular in a machine line for filling and packaging food and / or beverages, wherein the machine line comprises a plurality of machines, and wherein the system comprises: a server on which a large language model, LLM, is implemented, wherein a control command context and a specific context are defined for the LLM, wherein: the control command context comprises a set of control commands available for the machine line and their syntax, the specific context defines a restriction of an output of the LLM, and the specified restriction restricts at least one output format, structure, and / or information source used for a response from the LLM; and a user interface for interacting with the LLM; wherein the LLM is adapted to: Entering machine line status data; Receiving a user request via the user interface regarding the machine line; Processing the user request based on the specified control command context, the specific context and based on the state data; Creating a response to the user request based on the processing, the response comprising a control command generated by the LLM that is executable by the machine line, and Output the generated control command.

8. The system of claim 7, wherein the user interface is a web user interface, a voice assistant, or a computer application.

9. The system of claim 7 or 8, wherein the user interface is adapted to receive confirmation from a user that the control command is to be executed by the machine line.

10. The system of any one of claims 7 to 9, wherein the control command context further specifies one or more parameter value ranges for the set of control commands within which parameters may be changed by the control command.

Images