Computer-implemented method for automated configuration of a control system

Abstract

The present invention relates to a computer implementation method (100) for automated configuration of a control system (200), comprising: - collecting (102) first information (10) from at least a document database (60) by a document type identifier module (50), wherein the collection involves a step of detecting (103) document type information (61) of at least one document (62) found in at least one document database (60), and suggesting that the first information (10) be extracted from at least one document (62) having the identified document type information (61) if the identified document type information (61) of at least one document (62) satisfies a certain quality parameter (63); - extracting (104) the suggested first information (10) from at least one document (62) having the identified document type information (61) by an extraction module (52) to obtain second information (12); - generating (106) at least one factory segment (310) of a factory system (300) based on the second information (12) by a factory segment module (54); Configuration information (14) is generated (108) by the configuration module (55) for at least one control agent (220) of the control system (200) for the automation control of at least one plant section (310) of the plant system (300).

Description

Technical Field

[0001] This invention relates to a computer-based method for the automated configuration of a control system. Background Technology

[0002] The engineering design and configuration of automated systems (such as process equipment or industrial systems) is often a time-consuming task because the necessary information is typically distributed across multiple documents. Even experienced engineers must collect these documents and manually extract information. However, the workload remains significantly greater for inexperienced engineers, as they still need to figure out which information to look for in which documents. Once the information is collected, it must be translated into a control system architecture, which consists of well-configured control agents to automate or control the process equipment.

[0003] However, this part of the task is prone to error because it's easy for certain values ​​to not be correctly transmitted to the automated system. In any case, the configuration of today's automated systems still requires human intervention. As the field of automation strives for greater autonomy for several safety or efficiency-related reasons, the reconfiguration of automated systems that may become necessary while the system is running should not rely on human interaction. Therefore, a system capable of automatically configuring control systems is needed.

[0004] These problems need to be addressed. Summary of the Invention

[0005] Therefore, it would be advantageous to provide an improved concept for automatically generating configuration data for controlling factory systems in a highly efficient, safe, and automated manner.

[0006] The object of the invention is achieved by the subject matter of the independent claims, wherein further embodiments are included in the dependent claims.

[0007] In a first aspect of the present invention, a computer-based method for the automated configuration of a control system is provided, comprising the following steps: - The document type identifier module collects first information from at least one document database, wherein the collection involves the step of detecting document type information of at least one document found in at least one document database, and if the identified document type information of at least one document meets certain quality parameters, it is suggested that the first information be extracted from at least one document having the identified document type information. - The extraction module extracts suggested first information from at least one document with identified document type information to obtain second information; - The factory segment module generates at least one factory segment of the factory system based on the second information; - Configuration information is generated by the configuration module for at least one control agent of the control system for the automation control of at least one plant section of the plant system.

[0008] In other words, an important aspect of this invention is to automatically generate structured configuration data or configuration information for the control system based on available documentation of the process plant to be controlled by the control system.

[0009] This document includes various document types, such as P&ID diagrams, HMI diagrams, control narratives, function descriptions, and IO lists, which may vary depending on the specific application scenario of the system to be controlled.

[0010] According to the present invention, based on these documents, the factory system is divided into segments, and the control system is configured with at least one control agent per segment. The configuration file of the control agent includes input and output variables, as well as control limits, disturbance signals, rewards, objectives, etc.

[0011] Another important aspect of this invention is that a general information model (e.g., an AI model, such as LLM (Large Language Model)) is used to extract the information necessary for configuring and operating an automated control system that autonomously controls a factory system.

[0012] To achieve this, the pre-trained AI model can access process plant documents such as P&ID and HMI diagrams, control statements, function descriptions, I / O lists, naming conventions, equipment specifications, etc. The AI ​​model uses information found in these documents to divide the equipment into sections with minimal interaction. This information can be retrieved, for example, from HMI or P&ID diagrams. Naming conventions such as KKS (Power Plant Identification System) can also be used to identify plant sections. For each identified plant section, at least one control agent needs to be configured to control that plant section. In this case, the control agent corresponds to a piece of control software, such as an APC controller. Information includes input and output variables, controlled variables, targets, and control limits. Further information delivered by the AI ​​model can be a list of typical disturbance signals and failure modes for a given process.

[0013] In another aspect of the invention, one or more AI models are pre-trained to extract expected results from documents by using specially crafted prompts provided by the prompt builder module. This invention can be applied to greenfield applications (i.e., control systems configuring newly initiated processes) and brownfield applications (i.e., configuring or updating control agents for existing processes). The system also allows reconfiguration to be triggered by human experts or automatically.

[0014] According to the present invention, the automatically generated configuration information is used to establish an automation system for a new factory, but it can also be applied to reconfigure an already deployed automation system. In another aspect of the invention, the invented system and method can be triggered by personnel wishing to be supported in configuring an automation system, or invoked by another software system.

[0015] This invention provides an efficient solution for automated factory partitioning based on equipment and process documentation. To this end, an automated configuration of the factory partitioning control agent based on equipment and process documentation is generated. In the context of this invention, configuration refers to the specification of input and output variables, control limits, disturbance signals, etc.

[0016] Furthermore, according to the present invention, prompts are automatically generated to trigger the AI ​​model to search for information in a specific document type and to format the discovery in a usable manner.

[0017] This invention offers the following advantages, but is not limited thereto: - Reduced workload in configuring control systems. - Reduce human error when configuring control systems and process control equipment. - Increased productivity for control engineers. - Enabler for future agent-based autonomous control solutions. - Adaptive and automated (re)configuration of control agents for controlling plant systems.

[0018] According to the example, in the step of collecting the first information, at least a pre-trained general information model is used, which performs a search process to obtain the first information. In this way, the collection of the first information is performed in an efficient and automated manner.

[0019] According to the example, the general information model is pre-trained using at least one of the following training information as training data: information from similar devices, general information in process automation. In this way, the general information model can be configured and trained according to various application scenarios, and therefore it can be easily used in different applications.

[0020] According to the example, the general information model obtains document type-specific information or at least instructions that enable the general information model to identify specific document type information and / or output information about the expected result of a specific document type for at least one document through the prompting builder module. In this way, the general information model is trained in an efficient manner to obtain the desired output of the general information model.

[0021] According to the example, the document type-specific information provided by the prompt builder module includes at least one of the following: user-related input information, predefined template information, and checklist-related information. In this way, the training of the general information model is enriched by various data that improve the output of the general information model.

[0022] According to the example, the step of collecting the first information involves obtaining additional information, which comes from at least one of the following: prior knowledge, technical expert information. In this way, application-dependent information is used to improve the results of the extraction of the first information.

[0023] According to the example, document type information includes at least one of the following: a list of controlled variables, control limits of the variables, reward / target values ​​of the operation, typical operating conditions and known failure scenarios, and a P&ID diagram (i.e., piping and instrumentation diagram) to identify each plant section of the plant system. In this way, the document type information is enriched, thereby creating a more detailed data foundation for generating configuration information.

[0024] According to the example, the configuration information for at least one control agent includes at least one of the following: control objectives, operational constraints for the plant segment, and control variables that determine the operation of the plant system. In this way, effective control of the plant system is achieved.

[0025] According to the example, the generated configuration information is checked and verified by the verification module during the verification step to obtain verification information. This verification information is then provided to the configuration module via a feedback loop, causing the configuration information to be updated. In this way, the configuration information can be effectively adapted to the changing technical conditions and requirements of the factory system to be controlled.

[0026] As shown in the example, new input information from the document database is used to improve the pre-trained general information model. In this way, the general information model is enriched, thereby continuously improving its output.

[0027] In a second aspect of the invention, a control system is provided, which is configured to perform a method for automated configuration of a control system according to any of the foregoing examples and / or according to the first aspect.

[0028] In a third aspect of the invention, a computer is provided, including a processor configured to perform the method according to the first aspect and / or the method according to any of the foregoing examples.

[0029] In a fourth aspect of the invention, a computer program product is provided, including instructions that, when executed by a computer's processor, cause the computer to perform the method of the first aspect and / or any of the foregoing examples.

[0030] In a fifth aspect of the invention, a machine-readable data medium and / or downloadable product is provided, comprising a computer program according to the fourth aspect. Attached Figure Description

[0031] Exemplary embodiments will now be described with reference to the following figures:

[0032] Figure 1 A schematic flowchart illustrating the method of the present invention according to an embodiment of the present invention is shown;

[0033] Figure 2 A schematic flowchart illustrating the method of the present invention according to an embodiment of the present invention is shown; and

[0034] Figure 3 The illustration shows a schematic workflow diagram of a document type identifier module according to an embodiment of the present invention. Detailed Implementation

[0035] Figure 1 A schematic flowchart of the method 100 of the present invention according to an embodiment of the present invention is shown.

[0036] In the first step 102, the following is performed: the document type identifier module 50 collects first information 10 from at least one document database 60. The collection step involves step 103 of detecting document type information 61 of at least one document 62 found in at least one document database 60, and if the identified document type information 61 of at least one document 62 meets certain quality parameters 63, it is suggested that the first information 10 be extracted from the at least one document 62 having the identified document type information 61.

[0037] Optionally, document type information 61 includes at least one of the following: a list of controlled variables, control limits of the variables, reward / target values ​​of the operation, typical operating conditions and known failure scenarios, and P&ID diagrams (i.e., piping and instrumentation diagrams) for identifying each plant section 310, 320 of the plant system 300.

[0038] Optionally, in the step of collecting 102 first information 10, at least a pre-trained general information model 20 (e.g., an AI model) is used to perform a search process to obtain the first information 10.

[0039] Optionally, the general information model 20 is pre-trained using at least one of the following training information 22 as training data: information from similar devices, general information in process automation. Furthermore, the pre-trained general information model 20 can be improved by new input information 23 added to the document database 60.

[0040] Optionally, General Information Model 20—also refer to Figure 3—The prompt builder module 56 obtains document type-specific information 64 or at least instructions that enable the general information model 20 to identify specific document type information 25 and / or output information 24 regarding the expected result of a specific document type for at least one document 62. Optionally, the document type-specific information 64 provided by the prompt builder module 56 includes at least one of the following: user-related input information, predefined template information, and checklist-related information.

[0041] Optionally, step 102 of collecting the first information 10 involves obtaining additional information 26—see this document. Figure 2 —It comes from at least one of the following: prior knowledge 26-1, technical expert information 26-2 (see Figure 2 ).

[0042] In the second step 104, the following is performed: the extraction module 52 extracts the suggested first information 10 from at least one document 62 having identified document type information 61 to obtain the second information 12.

[0043] In the third step 106, the following is performed: at least one factory segment 310 of the factory system 300 is generated by the factory segment module 54 based on the second information 12. In the fourth step 108, the following is performed: configuration information 14 is generated by the configuration module 55 for at least one control agent 220 of the control system 200 for the automated control of at least one factory segment 310 of the factory system 300.

[0044] In the fourth step 108, the following is performed: configuration information 14 is generated by configuration module 55 for at least one control agent 220 of control system 200 for automation control of at least one factory section 310 of factory system 300.

[0045] Optionally, the configuration information 14 for at least one control agent 220 includes at least one of the following: control objectives, operational constraints for the plant segment, and control variables that determine the operation of the plant system 300.

[0046] Optionally, the generated configuration information 14 is checked and verified by the verification module 58 in verification step 110 to obtain verification information 59, wherein the verification information 59 of the verification process is provided to the configuration module 55 in a feedback-like manner (e.g., in feedback loop 65), so that the configuration information 14 is updated (see [link to documentation]). Figure 2 ).

[0047] Figure 2 A schematic workflow diagram of the method 100 of the present invention according to an embodiment of the present invention is illustrated. The present invention will be described in detail below by providing further aspects of possible implementations of the invention.

[0048] In order to obtain relevant second information 12 by extracting suggested first information 10 from the document database 60, one or more pre-trained AI models 20 that can access device and process documents 62 are used.

[0049] The AI ​​model queries document 62 as a process document, and generates configuration information or configuration file 14, factory segment 310, and control agent 220 of factory segment 310 responsible for factory system 300.

[0050] The prompt constructor 56 delivers prompts, document type-specific information, or simple instructions 64, which enable the general information model 20 to identify specific document type information 25 and / or output information 24 regarding the expected result for a specific document type of at least one document 62 (see [link]). Figure 2 and Figure 3 ).

[0051] Prompts are needed to generate the control system architecture for control system 200, which consists of multiple control agents 220. These control agents 220 control various sections 310 of the plant system or engineering system 300 through individual agent configurations (see [link]). Figure 2 ).

[0052] Typical documents used by AI model 20 include P&ID and HMI diagrams, control statements, IO lists, naming conventions, equipment specifications, etc. AI model 20 can also be enriched using general knowledge 26-1 in process automation, and can be pre-trained using information from similar past plants or input from technical experts 26-2. Figure 2 ).

[0053] Techniques such as one-shot / few-shot hints, where the hints to AI model 20 also include information about the expected outcome 24 for each document type, can be used to obtain the desired output from AI model 20 (see [link to relevant documentation]). Figure 2 ).

[0054] The prompts necessary to generate the desired output are automatically generated by the prompt builder module 56, which triggers the AI ​​model 20 to find the control system architecture for building the control system 200 and / or to generate the corresponding document information necessary for the plant segment 310 of the plant system 300 controlled by each control agent 220 using configuration information or the corresponding configuration file 14. Figure 2 and Figure 3 The prompt constructor module 56 is capable of generating special prompts that incorporate specific requirements, such as those derived from the reasons for the reconfiguration of the control system 200.

[0055] The proposed methods and systems can be triggered: - The establishment of the control system 200 is triggered manually. Manuals can use queries generated by the prompt builder module 56, modify them as needed, or generate their own queries. - Triggered by software components of the automatic configuration automation control system.

[0056] This invention can be used before the process plant is started or during the operation of the (process) plant system 300. The latter can occur, for example, when the plant topology changes or the current configuration proves to be suboptimal, such as due to frequent violations of control limits or unsatisfactory production performance. In this case, software components monitoring plant KPIs can trigger a reconfiguration. The generated configuration file 14 can be added to the document database 60 by the feedback loop 65 to further enhance the pre-trained AI model 20 for future use.

[0057] Now for reference Figure 2 The workflow of the proposed solution begins with identifying various document types 61 in the process document database 60. The document type identifier module 50 traverses the documents 62 in the database 60 to detect the document type information 61 of each document 62 and suggests that first information 10 (= second information 12) be extracted from the document.

[0058] Then, secondly, expected information 24 (see...) Figure 3 Together with prior knowledge 26-1, further information 26-2, and documents (or document fragments), it is passed to the formatted information extraction module 52 to extract information 12 that will be directly used to configure the control agent 220, including the plant section 310, the list of controlled / manipulated / measured variables, the control limits of the variables, and the rewards / goals of the operation.

[0059] Other types of information, such as typical operating conditions and known failure scenarios, can also be extracted and used when configuring agents. For example, multiple control agents 220 with different operational objectives can be configured based on different operating scenarios. One control agent 220 can be configured to optimize the productivity of a certain plant segment 310; another control agent can be configured to ensure that operations are within safety boundaries.

[0060] The extracted second information 12 can also be verified by using the output verification module 58 to reference various types of information, such as prior knowledge of similar processes, first-principles knowledge, and historical information of the target process. For example, control limits of process variables can be verified against variables from the same unit of a similar process to ensure that the control limits are not extreme. Only verified outputs are used to configure the control agent 220. Then, before using the extracted information 12 to configure the control agent 220, a checklist of information required by the control agent 220 is applied to check the completeness of the extracted information 12. If specific information is missing, a dedicated query module uses a dedicated query 67 (see...). Figure 2 This creates more specific prompts to retrieve the required information or clarify that the required information is unavailable. The confirmed configuration will be recorded in the document database 60 for future reference when updating the control agent configuration or configuring the control agent for other processes.

[0061] After the initial configuration of the control agent 220, it can be triggered again. Figure 2 The workflow within the framework generates new information for configuring and / or reconfiguring the control agent: - When new process documents are added to the document database 60, these documents will be automatically used to retrain or fine-tune the pre-trained AI model 20; workflows will also be triggered for new documents to identify whether additional information is included, and whether any agents need to be updated and how they are updated. - When the current performance of control agent 220 is no longer satisfactory, reconfiguration can be automatically triggered by, for example, an orchestrator that monitors the performance of control agent 220. - Workflows can also be triggered by human users, for example, if the current performance of control agent 220 is no longer satisfactory and a reconfiguration with new preferences should be triggered. - Part of this invention is establishing a configuration database for previously managed plants ( Figures 1 to 3 (Not shown in the image). In such a configuration database, configurations from the identified plant sections are stored, including the section's functional description, design and operating specifications, definitions of (input and output) control signals, control loops, alarms, interlocks, process topology, etc. It is possible to detect whether one or more process areas are missing from available documentation by comparing configuration results generated from similar equipment. For example, when processing available documentation from a power plant, it is easy to detect that information (complete or partial) about the condensing unit is unavailable. This can be done by comparing it with segments from other power plants available in the configuration database. Furthermore, using a dedicated query module 68, solutions that have already worked in other similar plants can be suggested.

[0062] In this context, and as a further example according to the invention, when certain information is missing in the configuration file, such as the target of a certain control agent, the missing information will be detected by using feedback loop 65, and the system will be instructed to generate the missing information.

[0063] The ability to compare and combine information contained in available documentation for existing plants with information and configurations of validated solutions in the configuration database makes it possible to check integrity and progressively improve the final configuration results for a specific plant.

[0064] In summary, the following aspects are relevant to this invention: - The proposed solution provides automated factory partitioning based on factory and process documentation. The proposed solution also provides automated configuration of plant partition control agents based on plant and process documentation. In this case, configuration refers to the specification of input and output variables, control limits, disturbance signals, etc. - Hint Builder module 56 automatically generates hints, which trigger the AI ​​model to find information in a specific document type and format the discovery in an appropriate manner.

[0065] Figure 3 A schematic workflow diagram of the document type identifier module 50 according to an embodiment of the present invention is illustrated. The prompt constructor module 56 constructs a prompt or document type-specific information 64 to query the type of a process document. The found document, along with the prompt, is given to a pre-trained AI model 20, such as a large language model. The AI ​​model 20 then provides specific document type information 25 to identify the type of the process document 62, as well as expected output information 24 that can be extracted from the document 62. A document may have more than one type of information. For example, a process's P&ID diagram may include both the label name and connectivity of process units; a control description may include controlled variables and their limits. In addition to the process document 62 itself, both the document type information 25 and the expected output information 24 are used by the (formatting information) extraction module 52 to extract formatting information from the document 62.

[0066] The workflow of the formatted information extraction module 52 can be described as follows: A hint 64 is constructed by combining multiple examples of questions / answers, document fragments related to a specific question, input from domain experts 26-2, and the expected format of output 24. The hint 64 is then fed to an AI model 20, such as a pre-trained large language model, which is capable of following the multiple examples and generating output in the expected format based on the multiple document fragments. The output is then validated by the validation module 58 and added to the document database 60 via a feedback information stream 59 for use in controlling the configuration of the agent 220. Figure Labels 10 First Information 12 Second Information 14 Configuration Information / Configuration Files 20 General Information Model 22 Training Information 23 Input Information 24 Output / Result Information 25. Specific document type information 26 External / Additional Information 26-1 Prior Knowledge 26-2 Technical Expert Information 28 Checklist Information 50 Document Type Identifier Module 52 (Formatted Information) Extraction Module 54 Factory Section Module 55 Configuration Module 56 Hints on constructor modules 58. Verification Module 59 Verification Message Results 60 (Document) Database 61. Document Type Information 62 documents 63 Quality Parameters 64. Document type-specific information / instructions 65 Feedback Loop 66 Configure Control Agent 67 Query 68 Query Module 100 Computer Implementation Methods 102 Collections 103 Detection 104 Extraction 106 generated 108 generated 110 verification 200 Control System 220 Control Agent 300 Factory Systems / Engineering Systems 310 Factory Section

Claims

1. A computer-based method (100) for automated configuration of a control system (200), comprising: The document type identifier module (50) collects (102) first information (10) from at least one document database (60), wherein the collection involves the step of detecting (103) document type information (61) of at least one document (62) found in the at least one document database (60), and if the identified document type information (61) of the at least one document (62) meets a certain quality parameter (63), it is suggested that the first information (10) be extracted from the at least one document (62) having the identified document type information (61); The extraction module (52) extracts (104) the first information (10) suggested from the at least one document (62) having the identified document type information (61) to obtain the second information (12). At least one factory segment (310) of the factory system (300) is generated (106) by the factory segment module (54) based on the second information (12). Configuration information (14) is generated (108) by the configuration module (55) for at least one control agent (220) of the control system (200) for the automation control of at least one factory section (310) of the factory system (300).

2. The computer implementation method (100) according to claim 1, wherein in the step of collecting (102) the first information (10), at least a pre-trained general information model (20) is used, the pre-trained general information model (20) performing a search process to obtain the first information (10).

3. The computer implementation method (100) according to claim 2, wherein the general information model (20) is pre-trained by using at least one of the following training information (22) as training data: information from similar devices, general information in process automation.

4. The computer implementation method (100) according to claim 2, wherein the general information model (20) obtains document type-specific information (64) or at least instructions that enable the general information model (20) to identify specific document type information (25) and / or output information (24) of the expected result of the specific document type of the at least one document (62) through the prompt builder module (56).

5. The computer implementation method (100) according to claim 4, wherein the document type-specific information (64) provided by the prompt builder module (56) includes at least one of the following: user-related input information, predefined template information, and checklist-related information.

6. The computer implementation method (100) according to any one of the preceding claims, wherein the step (102) of collecting the first information (10) involves obtaining additional information (26), said additional information (26) being derived from at least one of the following: prior knowledge (26-1) and technical expert information (26-2).

7. The computer implementation method (100) according to any one of the preceding claims, wherein the document type information (61) includes at least one of the following: a list of controlled variables, control limits of the variables, reward / target values ​​of the operation, typical operating conditions and known failure scenarios, and a P&ID diagram (i.e., piping and instrumentation diagram) for identifying each plant section (310, 320) of the plant system (300).

8. The computer implementation method (100) according to any one of the preceding claims, wherein the configuration information (14) for the at least one control agent (220) includes at least one of the following: control objectives, operational constraints for the factory section, and control variables for determining the operation of the factory system (300).

9. The computer implementation method (100) according to any one of the preceding claims, wherein the generated configuration information (14) is checked and verified by a verification module (58) in a verification step (110) to obtain verification information (59), wherein the verification information (59) of the verification process is provided to the configuration module (55) according to a feedback loop (65) so that the configuration information (14) is updated.

10. The computer implementation method (100) according to any one of claims 2 to 9, wherein new input information (23) in the document database (60) is used to improve the pre-trained general information model (20).

11. A control system (200) configured to perform a method for automated configuration of a control system according to any one of the preceding claims.

12. A computer comprising a processor configured to perform the method of any one of claims 1 to 10.

13. A computer program product comprising instructions that, when executed by a processor of a computer, cause the computer to perform the method of any one of claims 1 to 10.

14. A machine-readable data medium and / or downloadable product comprising the computer program according to claim 13.

Images