Method for Interactive Improving Manufacturing Process for Manufacturing System Based on Large Language Model

Abstract

The present invention relates to a method for improving manufacturing processes in a manufacturing system based on a large language model, and more specifically, to a method for improving manufacturing processes in a large language model-based manufacturing system that maximizes response speed and operational efficiency at the manufacturing site by linking equipment data accumulated in a database with schedule data from an APS, wherein the large language model interprets natural language-based user requests and determines whether to perform the process, and infers the cause of non-compliance with the schedule to generate and update an optimized improvement schedule.

Description

Technology Field

[0001] The present invention relates to a communicative manufacturing process improvement method for a manufacturing system based on a large language model, and more specifically, to a communicative manufacturing process improvement method for a large language model-based manufacturing system that maximizes response speed and operational efficiency at the manufacturing site by linking equipment data accumulated in a database with schedule data from Advanced Planning and Scheduling (APS), wherein a large language model (LLM) interprets natural language-based user requests and determines whether to perform a process, and infers the cause of non-compliance with the schedule to generate and update an optimized improvement schedule. Background Technology

[0003] As modern manufacturing shifts toward a multi-product, small-batch production system, the complexity of manufacturing processes is rapidly increasing, leading to an accelerated adoption of smart factories to address this. Generally, smart factories operate based on Advanced Planning and Scheduling (APS), which establishes production plans, and Manufacturing Execution Systems (MES), which collect and control real-time data from the manufacturing site. APS generates work schedules considering resource constraints and delivery deadlines, while MES manages historical data and sensing information, such as equipment operation, shutdown, and production volume, accordingly.

[0004] However, conventional manufacturing systems suffer from the problem that APS and MES are disconnected and not organically integrated. For example, even if the process proceeds according to the schedule established by the APS, it is difficult for the APS to immediately detect and modify the schedule when unexpected situations, such as equipment failures or material shortages, occur on-site. Typically, the system has an inefficient structure where a manager must detect an anomaly via the MES monitoring screen, manually access the APS to change the plan, or issue verbal instructions to on-site workers.

[0005] Furthermore, existing monitoring systems are limited to merely listing equipment status codes or numerical data. Consequently, it is impossible for the system to understand and respond when an operator asks contextual questions such as, "Why has equipment C stopped?" or "Is it possible to achieve the target production volume?" Since operators must manually query and analyze scattered data, it takes a significant amount of time to identify the cause of a problem and devise countermeasures unless one is a skilled expert; this ultimately leads to increased downtime and decreased productivity across the entire process.

[0006] Furthermore, the lack of standardization in equipment data makes integrated analysis between heterogeneous equipment difficult, and relying on fragmentary rule-based logic when establishing improvement schedules for problem solving limits the ability to derive optimal scenarios that consider complex variables.

[0007] Therefore, there is an urgent need to develop a communicative manufacturing process improvement method that utilizes Large Language Models (LLMs) capable of natural language processing to understand user intent, integrates and analyzes data from MES and APS to infer the causes of problems, and further automatically generates actionable improvement schedules to reflect in the system. Prior art literature

[0009] Korean Registered Patent KR 10-2064490 B1 “Manufacturing and Production Process Management System” (2020.01.03) The problem to be solved

[0010] The present invention aims to provide a method for improving manufacturing processes in a manufacturing system based on a large language model, and more specifically, to provide a method for improving manufacturing processes in a large language model-based manufacturing system that maximizes response speed and operational efficiency at the manufacturing site by linking equipment data accumulated in a database with schedule data from an APS, allowing the large language model to interpret natural language-based user requests, determine whether to perform the process, infer the cause of non-compliance with the schedule, and generate and update an optimized improvement schedule. means of solving the problem

[0012] To solve the above-mentioned problem, a communicative manufacturing process improvement method for a large language model-based manufacturing system is performed on a server system comprising one or more processors and one or more memories, wherein the server system includes in a database equipment data including history data and sensing data for each of a plurality of equipment included in the manufacturing system, received from a Manufacturing Execution System (MES) and an equipment monitoring system for the manufacturing system, and the manufacturing process improvement method comprises: a request reception step of receiving a user request in natural language form from a user terminal; and a data collection step of receiving schedule data related to the user request from an Advanced Planning and Scheduling (APS) for the manufacturing system and collecting related equipment data including equipment data for each of all equipment related to the schedule data from the database. A method for improving a manufacturing process through communication is provided, comprising: an operation judgment step of inputting the above user request, the above schedule data, and the above related equipment data into an internal or external large language model to derive operation judgment information for each of a plurality of equipment related to the above schedule data, including whether the equipment has normally performed the operation that the equipment is required to perform according to the above schedule data; an improvement schedule generation step of generating re-established improvement schedule data based on equipment data for the equipment, the above schedule data, and the above user request, so that the equipment achieves a target production volume or target completion time according to the above schedule data, when equipment that has not normally performed the operation corresponding to the above schedule data is identified in the operation judgment information; and a schedule update step of transmitting the improvement schedule data to the above APS to cause the manufacturing system to perform process operations according to the improvement schedule data.

[0013] In one embodiment of the present invention, the improvement schedule generation step may include: a step of deriving the cause of non-operation of the non-operational equipment by inputting a prompt containing equipment data for the non-operational equipment into a large language model when a non-operational equipment corresponding to equipment that has not normally performed an operation corresponding to the schedule data is identified in the operation judgment information; and a step of deriving the improvement schedule data by inputting the relevant equipment data containing equipment data for each of all equipment related to the schedule data including the non-operational equipment, and a prompt containing the cause of non-operation into a large language model.

[0014] In one embodiment of the present invention, the improvement schedule generation step calculates the expected production volume when the manufacturing system operates according to the improvement schedule data, verifies the validity of the improvement schedule data based on whether the expected production volume is greater than or equal to the existing production volume according to the schedule data, and the schedule update step can transmit the improvement schedule data to the APS only when the validity of the improvement schedule data has been verified.

[0015] In one embodiment of the present invention, the user request includes a production quantity verification request inquiring whether the actual production quantity produced by the manufacturing system according to the schedule data has achieved the target production quantity that the manufacturing system should have produced when following the schedule data, and the operation judgment step may include: a step of calculating the target production quantity according to the schedule data and calculating the actual production quantity by receiving history data for each of a plurality of facilities related to the schedule data when the user request corresponds to a production quantity verification request; and a step of determining whether the target production quantity has been achieved by comparing the target production quantity with the actual production quantity.

[0016] In one embodiment of the present invention, the improvement schedule generation step may further include: a step of identifying a production-decreasing facility that, for each of a plurality of facilities related to the schedule data, has produced less than a predetermined threshold value than the production quantity that the facility should have produced according to the schedule data when the actual production quantity is less than the target production quantity; a step of inputting a prompt containing facility data for the production-decreasing facility into a large language model to derive the cause of the production-decreasing facility; and a step of inputting a prompt containing the schedule data, related facility data for each of all facilities related to the schedule data including the production-decreasing facility, and the cause of the production-decreasing facility into a large language model to derive the improvement schedule data.

[0017] In one embodiment of the present invention, the server system stores data in the form of a table including a plurality of rows and columns, and defines a mapping rule that defines which column of the equipment data for each of the plurality of equipment corresponds to which column of the integrated data.

[0018] The above database may store integrated data that is created or updated by integrating the equipment data for each of the plurality of equipment into one, while inputting the column-specific data values ​​of the equipment data for each of the plurality of equipment into the corresponding columns of integrated data according to the mapping rule.

[0019] In one embodiment of the present invention, the data collection step inputs the integrated data, in which the equipment data for each of the plurality of equipment is managed integrally, and the schedule data into a large language model, derives a related area in the integrated data where data related to the schedule data is stored, and extracts only the data corresponding to the related area from the integrated data to receive detailed integrated data.

[0020] In one embodiment of the present invention, the operation determination step is performed in an operation determination unit included in the server system, and the operation determination unit may include one or more of: a process monitoring unit that monitors a real-time process state based on sensing data included in the equipment data; and a state abnormality cause derivation unit that derives a cause corresponding to whether there is a mechanical defect, tool wear, or abnormal process condition that caused the state abnormality for the equipment where the state abnormality occurred.

[0021] To solve the above-mentioned problem, a server system comprising one or more processors and one or more memories, and performing a communicative manufacturing process improvement method for a large language model-based manufacturing system, wherein the server system stores equipment data including history data and sensing data for each of a plurality of equipment included in the manufacturing system, received from a Manufacturing Execution System (MES) and an equipment monitoring system for the manufacturing system; a request receiving unit that receives a user request in natural language form from a user terminal; a data collection unit that receives schedule data related to the user request from an Advanced Planning and Scheduling (APS) for the manufacturing system and collects related equipment data related to the schedule data from the database; and an operation judgment unit that inputs the user request, the schedule data, and the related equipment data into an internal or external large language model to derive operation judgment information for each of a plurality of equipment related to the schedule data, including whether the equipment has normally performed the operation that it is required to perform according to the schedule data. A server system is provided comprising: an improvement schedule generation unit that, when equipment that has not normally performed an operation corresponding to the schedule data is identified in the operation judgment information, generates improved schedule data re-established so that the equipment achieves a target production volume or target completion time according to the schedule data based on equipment data for the equipment, the schedule data, and the user request; and a schedule update unit that transmits the improved schedule data to the APS to cause the manufacturing system to perform process operations according to the improved schedule data. Effects of the invention

[0023] According to one embodiment of the present invention, by receiving a request in the form of natural language from a user and interpreting it to perform a linked analysis of the schedule data of the APS and the equipment data of the MES, even an operator without specialized knowledge can easily grasp and control the process status.

[0024] According to one embodiment of the present invention, the operation of equipment is automatically determined based on schedule data, and when non-operating equipment is identified, a large language model infers the cause of non-operation based on the equipment data of the equipment and generates an improved schedule, thereby enabling a rapid and accurate response in the event of a process anomaly.

[0025] According to one embodiment of the present invention, by calculating the expected production volume based on the improvement schedule data, verifying it by comparison with the existing target production volume, and updating the schedule only when valid, the reliability of the improved schedule can be secured and the process stability can be maintained.

[0026] According to one embodiment of the present invention, when a request is made to verify production volume, the target production volume and the actual production volume are compared and analyzed, and if the target is not met, the equipment causing the production decline is identified and the cause is specifically presented, thereby enabling the early elimination of factors causing reduced productivity and the improvement of process yield.

[0027] According to one embodiment of the present invention, by deriving improvement schedule data based on the results of the cause analysis of production degradation equipment, it is possible to achieve the effect of providing a practical solution to the problem beyond simply identifying the phenomenon.

[0028] According to one embodiment of the present invention, by normalizing various data from heterogeneous equipment according to mapping rules and managing them as integrated data, it is possible to achieve the effect of building a scalable system regardless of the type of equipment.

[0029] According to one embodiment of the present invention, by selectively extracting only the relevant areas necessary for schedule analysis from vast integrated data and inputting them into a large language model, it is possible to achieve the effect of increasing data processing speed and reducing computation costs.

[0030] According to one embodiment of the present invention, through a dedicated module that performs real-time process monitoring and identifies the cause of abnormal conditions, signs of physical abnormalities such as mechanical defects or tool wear can be immediately detected, thereby enabling predictive maintenance.

[0031] According to one embodiment of the present invention, scenarios for a plurality of improved schedules reflecting user requirements are generated, and an optimal alternative is presented by calculating priorities based on expected production volume and process time, thereby effectively supporting the manager's decision-making. Brief explanation of the drawing

[0033] FIG. 1 illustrates the configuration of a server system that performs a communication-type manufacturing process improvement method according to one embodiment of the present invention. FIG. 2 illustrates the process of performing the steps and schedule update steps of a communication-type manufacturing process improvement method according to one embodiment of the present invention. FIG. 3 illustrates a process of deriving operation judgment information in an operation judgment step according to an embodiment of the present invention. FIG. 4 illustrates an example of deriving improvement schedule data when non-operational equipment is identified in the improvement schedule generation step according to an embodiment of the present invention. FIG. 5 illustrates a process for verifying the validity of improvement schedule data according to one embodiment of the present invention. FIG. 6 illustrates a process of determining whether the target production amount is achieved by calculating the target production amount and the actual production amount in the operation judgment step when a user request according to an embodiment of the present invention corresponds to a request to verify the production amount. FIG. 7 illustrates the process of identifying production-degrading equipment and deriving improvement schedule data in the improvement schedule generation step according to one embodiment of the present invention. FIG. 8 illustrates the process of deriving the final improvement schedule data among a plurality of improvement schedule data generated by reflecting user requirements in the improvement schedule generation step according to an embodiment of the present invention. FIG. 9 illustrates the configuration of a database according to one embodiment of the present invention. FIG. 10 illustrates the configuration of integrated data stored in a database according to one embodiment of the present invention. FIG. 11 illustrates a process of deriving detailed integrated data containing only data related to schedule data in a data collection step according to an embodiment of the present invention. FIG. 12 illustrates the process of updating integrated data according to one embodiment of the present invention. FIG. 13 illustrates the configuration of a sub-unit included in an operation judgment unit according to one embodiment of the present invention. FIG. 14 schematically illustrates the internal configuration of a computing device according to one embodiment of the present invention. Specific details for implementing the invention

[0034] Hereinafter, various embodiments and / or aspects are disclosed with reference to the drawings. For illustrative purposes, numerous specific details are disclosed in the following description to aid in a general understanding of one or more aspects. However, it will also be recognized by those skilled in the art that these aspects may be practiced without such specific details. The following description and the accompanying drawings describe specific exemplary aspects of one or more aspects in detail. However, these aspects are exemplary, and some of the various methods in the principles of the various aspects may be used, and the description is intended to include all such aspects and their equivalents.

[0036] In addition, various aspects and features will be presented by a system that may include multiple devices, components and / or modules, etc. It should also be understood and recognized that various systems may include additional devices, components and / or modules, etc., and / or may not include all of the devices, components, modules, etc. discussed in relation to the drawings.

[0037] Terms such as “embodiment,” “example,” “aspect,” “example,” etc. as used herein may not be interpreted as implying that any aspect or design described is superior or more advantageous than other aspects or designs. Terms used below, such as “part,” “component,” “module,” “system,” “interface,” etc., generally refer to computer-related entities and may, for example, refer to hardware, a combination of hardware and software, or software.

[0038] Additionally, the terms “comprising” and / or “comprising” should be understood to mean that the relevant feature and / or component is present, but not to exclude the presence or addition of one or more other features, components and / or groups thereof.

[0039] Additionally, terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but said components are not limited by said terms. Such terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. The term "and / or" includes a combination of a plurality of related described items or any of a plurality of related described items.

[0040] Furthermore, in the embodiments of the present invention, all terms used herein, including technical or scientific terms, unless otherwise defined, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in the embodiments of the present invention.

[0042] FIG. 1 illustrates the configuration of a server system that performs a communication-type manufacturing process improvement method according to one embodiment of the present invention.

[0043] A method for improving a manufacturing process in a large language model-based manufacturing system according to one embodiment of the present invention is performed on a server system comprising one or more processors and one or more memories, wherein the server system includes in a database equipment data comprising history data and sensing data for each of a plurality of equipment included in the manufacturing system, received from a Manufacturing Execution System (MES) and an equipment monitoring system for the manufacturing system, and the method for improving the manufacturing process comprises: a request receiving step of receiving a user request in the form of natural language from a user terminal; and a data collection step of receiving schedule data related to the user request from an Advanced Planning and Scheduling (APS) for the manufacturing system, and collecting related equipment data including equipment data for each of all equipment related to the schedule data from the database. The method may include: an operation judgment step of inputting the above user request, the above schedule data, and the above related equipment data into an internal or external large language model to derive operation judgment information for each of a plurality of equipment related to the above schedule data, including whether the equipment has normally performed the operation that it is required to perform according to the above schedule data; an improvement schedule generation step of generating re-established improvement schedule data based on equipment data for the equipment, the above schedule data, and the above user request, so that the equipment achieves the target production volume or target completion time according to the above schedule data, when equipment that has not normally performed the operation corresponding to the above schedule data is identified in the operation judgment information; and a schedule update step of transmitting the improvement schedule data to the above APS to cause the manufacturing system to perform process operations according to the improvement schedule data.

[0045] As illustrated in FIG. 1, a server system (1) according to one embodiment of the present invention can be connected to a manufacturing system (2) and a user terminal (3) to transmit and receive data.

[0046] The above manufacturing system (2) may include a plurality of facilities at the actual product production site, an APS (20) that stores production plans, an MES (21) that collects and stores facility data including history data and sensing data for the plurality of facilities at the manufacturing site, and a facility monitoring system (not shown).

[0047] The above APS (20) corresponds to a system that plans and stores the sequence and schedule of work considering material requirements or production capacity, and the above MES (21) and equipment monitoring system corresponds to a system that manages and stores real-time data including sensing data generated from equipment at the site and work history.

[0048] The above server system (1) may include a request receiving unit (10), a data collection unit (11), an operation judgment unit (12), an improvement schedule generation unit (13), a schedule update unit (14), and a database (15).

[0049] The above request receiving unit (10) can perform a request receiving step (S10) of receiving a user request including a query or command in natural language text or voice from a user terminal (3).

[0050] The data collection unit (11) may perform a data collection step (S11) based on the user request, receiving schedule data containing information on a production plan from the APS (20), and collecting related equipment data including equipment data for a plurality of equipment related to the schedule data by querying equipment data for a plurality of equipment stored in the database (15). At this time, the equipment data may include the operation history of the equipment, alarm logs, sensor values, etc.

[0051] The above operation judgment unit (12) can perform an operation judgment step (S12) by inputting the collected schedule data and the related equipment data into a large language model inside or outside the server system to derive operation judgment information including whether the equipment actually operated according to the planned schedule.

[0052] Based on the above operation judgment information, if there is equipment that operates differently from the schedule or has stopped, the improvement schedule generation unit (13) can derive the cause based on the equipment data for the equipment, and input the cause, the above schedule data, and the above related equipment data into a large language model to generate a new schedule, i.e., improvement schedule data, which can solve or bypass the problem.

[0053] The above schedule update unit (14) transmits the improvement schedule data back to the APS (20) to perform a schedule update step (S14) for updating the production plan, thereby enabling the server system (1) to control the manufacturing process so that it is optimized according to real-time field conditions.

[0055] The above database (15) constructs integrated data based on mapping rules to efficiently manage data of multiple facilities included in the manufacturing system (2), and a specific explanation of this will be described later in the description related to FIGS. 9 and 10.

[0057] According to one embodiment of the present invention, by receiving a request in the form of natural language from a user and interpreting it to perform a linked analysis of the schedule data of the APS and the equipment data of the MES, even an operator without specialized knowledge can easily grasp and control the process status.

[0058] In addition, the flexibility and responsiveness of the manufacturing system can be significantly improved by actively correcting manufacturing plans through large language models.

[0060] FIG. 2 illustrates the process of performing the steps and schedule update steps of a communication-type manufacturing process improvement method according to one embodiment of the present invention.

[0062] As illustrated in FIG. 2(a), the manufacturing process improvement method of the present invention includes a request reception step (S10), a data collection step (S11), an operation determination step (S12), and a schedule improvement step (S13), and the server system (1) can perform each step sequentially.

[0064] First, the request reception step (S10) is performed in the request reception unit (10), and the request reception step (S10) can receive a user request including a natural language question entered as text or voice from a worker terminal.

[0066] The subsequent data collection step (S11) is performed in the data collection unit (11), and the schedule data related to the user request can be collected from the APS (20) by extracting key keywords such as equipment name, process name, and date included in the user request to query the APS (20), or by inputting one or more schedule data stored in the user request and the APS (20) into a large language model.

[0067] In addition, related equipment data including equipment data for a plurality of equipment included in the schedule data can be collected by accessing the database (15). Specific details for deriving the related equipment data from the database (15) will be described later in the description related to FIG. 11.

[0069] The above-mentioned related facility data includes only the facility data for the plurality of facilities included in the schedule data among the facility data for the plurality of facilities stored in the above-mentioned database.

[0070] For example, when there are equipment A, B, C, and D in the above manufacturing system (1), if the schedule data is “Equipment A → Equipment B → Equipment C → Production of Product X”, the related equipment data may include equipment data for equipment A, B, and C, but not equipment data for equipment D.

[0071] In other words, inputting related facility data (Facility A, B, C) into a large language model and inputting facility data for multiple facilities (Facility A, B, C) related to schedule data into a large language model can be considered equivalent.

[0072] Furthermore, inputting related equipment data (equipment A, B, C) into a large language model and inputting equipment data of a specific piece of equipment (which may include equipment A, B, C, or D) into a large language model may have different meanings.

[0074] The above operation judgment step (S12) is performed in the operation judgment unit (12), and the user request, the schedule data, and the related equipment data can be input into an internal or external large language model. The large language model compares and analyzes the planned work instructions with the actual equipment operation logs to determine whether each piece of equipment is operating normally as planned, or whether it is stopped or delayed, and can derive operation judgment information from the result.

[0076] If equipment that has not normally performed the operation corresponding to the schedule data is identified based on the above operation judgment information, the above improvement schedule generation step (S13) can be performed by the above improvement schedule generation unit (13).

[0077] The above improvement schedule generation step (S13) can derive equipment data for equipment that has not normally performed the operation corresponding to the schedule data from the database (15) or the related equipment data.

[0078] In addition, one or more of the equipment data, the schedule data, the related equipment data, and the user request can be input into a large language model to generate improved schedule data that is re-established to achieve the target production volume or target completion time according to the schedule data.

[0079] In other words, by identifying problematic equipment and considering information about the equipment, the existing production plan, information about multiple pieces of equipment related to the existing production plan, and user requests, an improved plan capable of achieving the goals of the existing production plan can be derived.

[0081] Finally, the schedule update step (S14) is performed by the schedule update unit (14), and as shown in FIG. 2 (b), the improvement schedule data generated in the improvement schedule generation step (S13) is transmitted to the APS (20) based on the equipment data, related equipment data, schedule data, and user request received from the database (15), the user terminal (3), and the APS (20), respectively, thereby updating the existing schedule data into the improvement schedule data. Thus, the manufacturing system (2) can immediately modify equipment control commands and perform subsequent process operations according to the updated schedule.

[0083] FIG. 3 illustrates a process of deriving operation judgment information in an operation judgment step according to an embodiment of the present invention.

[0084] Specifically, FIG. 3(a) illustrates an example of data that serves as input to the operation judgment step (S12), and FIG. 3(b) illustrates the flow in which the operation judgment information is derived through a large language model.

[0086] As illustrated in FIG. 3(a), the input data for the operation judgment step (S12) may consist of the schedule data, the related equipment data, and the user request.

[0087] The above schedule data is manufacturing plan information received from the APS (20), and may include, for example, a process flow such as "Facility A → Facility B → Facility C → Product X production", the start and end times of work for each facility, and the target production quantity.

[0089] The above user request is a request in the form of natural language entered by the user, and may include content asking whether the process is proceeding normally, such as "Is there a problem with Schedule X?"

[0091] As illustrated in FIG. 3(b), in the operation judgment step (S12), the schedule data, the related equipment data, and the user request can be input into a large language model inside or outside the server system (1). Through this, the large language model can derive operation judgment information including whether each of the plurality of equipment included in the schedule data has successfully performed an operation corresponding to the schedule data.

[0092] For example, if an error code is found in the equipment data of Equipment B included in the aforementioned related equipment data at the time when Equipment B is scheduled to operate, the large language model can determine this as an abnormal situation. Including whether each piece of equipment is performing normally, the operation judgment information may further include specific phenomena occurring when not performing and anticipated problems.

[0094] FIG. 4 illustrates an example of deriving improvement schedule data when non-operational equipment is identified in the improvement schedule generation step according to an embodiment of the present invention.

[0095] The improvement schedule generation step according to one embodiment of the present invention may include: a step of deriving the cause of non-operation of the non-operational equipment by inputting a prompt containing equipment data for the non-operational equipment into a large language model when a non-operational equipment corresponding to equipment that has not normally performed an operation corresponding to the schedule data is identified in the operation judgment information; and a step of deriving the improvement schedule data by inputting the relevant equipment data containing equipment data for each of all equipment related to the schedule data including the non-operational equipment, and a prompt containing the cause of non-operation into a large language model.

[0097] Figure 4(a) illustrates a situation in which equipment B is identified as non-operational equipment by the operation judgment information above.

[0098] In the improvement schedule generation step (S13), it can be determined whether non-operating equipment that has not normally performed the operation corresponding to the schedule data is identified based on the operation judgment information derived from the operation judgment step (S12). In the example of FIG. 4 (a), Equipment B was identified as the non-operating equipment in the operation judgment information based on the schedule data "Equipment A → Equipment B → Equipment C → Product X production".

[0099] In the above improvement schedule generation step (S13), the cause of why equipment B did not operate normally is identified, and improvement schedule data can be derived based on the cause.

[0101] Figure 4(b) illustrates the process of deriving non-operation cause and improvement schedule data through a large language model.

[0102] When the above non-operating equipment is identified, the equipment data of the above non-operating equipment can be collected from the above database.

[0103] In the improvement schedule generation step (S13), a prompt including one or more of the equipment data of the non-operating equipment and the related equipment data is input into a large language model to derive the cause of non-operation corresponding to the reason why the non-operating equipment did not operate normally. For example, causes such as "emergency stop due to motor overheating" or "network communication error" may be derived as the cause of non-operation.

[0104] In the example of FIG. 4(b), the cause of non-operation can be derived by inputting the equipment data of equipment B into a large language model as the equipment data of the non-operational equipment. In addition, the related equipment data, which includes equipment data for multiple equipment included in the schedule data, namely the equipment data of equipment A, equipment B, and equipment C, can be input into the large language model.

[0105] Next, in the improvement schedule generation step (S13), the derived cause of non-operation, the schedule data, and related equipment data can be input back into a large language model to derive the improvement schedule data. Based on the input data, the large language model can derive alternatives to resolve or bypass the cause of non-operation.

[0106] For example, as shown in the example of (c) in Fig. 4, if it is determined that repair of equipment B is not possible within a short period of time, spare equipment D that is idle can be introduced, or the manufacturing sequence can be readjusted to skip equipment B and proceed to the possible processes.

[0108] According to one embodiment of the present invention, rather than simply changing the schedule according to mechanical rules, a large language model understands the specific cause of the equipment's non-operation and presents a context-appropriate solution, thereby enabling the effect of minimizing production disruptions and responding flexibly even in the event of a sudden equipment failure. In addition, cause analysis and the establishment of alternatives are automated, which can drastically reduce the decision-making time of a manager.

[0110] FIG. 5 illustrates a process for verifying the validity of improvement schedule data according to one embodiment of the present invention.

[0111] According to one embodiment of the present invention, the improvement schedule generation step calculates the expected production volume when the manufacturing system operates according to the improvement schedule data, verifies the validity of the improvement schedule data based on whether the expected production volume is greater than or equal to the existing production volume according to the schedule data, and the schedule update step can transmit the improvement schedule data to the APS only when the validity of the improvement schedule data has been verified.

[0113] As illustrated in FIG. 5, in the improvement schedule generation step (S13), the generated improvement schedule data is not sent directly to the APS (20) for updating, but an additional verification process can be performed.

[0114] First, in the improvement schedule generation step (S13), a simulation can be performed based on the improvement schedule data, or a prediction model can be used to calculate the expected production volume when the process is carried out according to the schedule.

[0115] Specifically, performing the above simulation may correspond to a process of deriving the production capacity by inputting variables such as process sequence, input equipment, and work time included in the improvement schedule data into a model that implements a virtual manufacturing environment, and simulating the situation in which the actual process is carried out.

[0116] In addition, using the above prediction model may mean a process of inputting the above improvement schedule data into a machine learning-based model that has learned historical data (equipment utilization rate, failure frequency, work time, etc.) to probabilistically infer the production volume under the corresponding conditions.

[0118] Next, the above-mentioned estimated production quantity is compared with the existing production quantity or target value defined in the existing schedule data; if the estimated production quantity is greater than or equal to the existing production quantity, or is within a preset error range acceptable to the user, the corresponding improvement schedule data can be determined to be 'valid' (YES).

[0119] If it is determined to be valid, in the schedule update step (S14), the improvement schedule data can be transmitted to the APS (20) to update the process plan.

[0120] Conversely, if the estimated production volume falls short of the standard (NO), the corresponding improvement schedule data is determined to be 'invalid' and may be discarded, or feedback may be provided to the large language model to request its regeneration.

[0122] According to one embodiment of the present invention, by calculating the expected production volume based on the improvement schedule data, verifying it by comparison with the existing target production volume, and updating the schedule only when valid, the reliability of the improved schedule can be secured and the process stability can be maintained.

[0124] FIG. 6 illustrates a process of determining whether the target production amount is achieved by calculating the target production amount and the actual production amount in the operation judgment step when a user request according to an embodiment of the present invention corresponds to a request to verify the production amount.

[0125] According to one embodiment of the present invention, the user request includes a production quantity verification request inquiring whether the actual production quantity produced by the manufacturing system according to the schedule data has achieved the target production quantity that the manufacturing system should have produced when following the schedule data, and the operation judgment step may include: a step of calculating the target production quantity according to the schedule data and receiving history data for each of a plurality of facilities related to the schedule data and calculating the actual production quantity when the user request corresponds to a production quantity verification request; and a step of determining whether the target production quantity has been achieved by comparing the target production quantity with the actual production quantity.

[0127] As illustrated in FIG. 6(a), the user may input a user request including a production quantity verification request, such as "Have you achieved the target production quantity?" through the user terminal (3). Specifically, the production quantity verification request may correspond to a natural language query that includes the user's intention to verify the execution performance of the manufacturing system (2) against the plan based on a specific period or the current time.

[0129] As illustrated in Fig. 6(b), in the operation judgment step (S12), when a user request including a request to verify the production quantity is received, the target production quantity and the actual production quantity can be calculated.

[0130] The above target production quantity is a value derived based on the schedule data received from the APS (20), and may correspond to the planned cumulative number of products produced or the number of process completions that should be achieved at the current time (or the time requested by the user) if the schedule included in the schedule data is executed normally without any problems.

[0131] The above actual production volume can be calculated based on equipment data including history data and sensing data for multiple facilities stored in the database (15). This may correspond to the number of products or the number of process completions that have been physically completed by the equipment at the manufacturing site.

[0133] Afterward, the above target production amount and the above actual production amount are compared, and if the actual production amount is greater than or equal to the target production amount (NO), the above server system (1) can generate a positive response to the user, such as “The goal has been achieved.”

[0134] On the other hand, if the target production amount is greater than the actual production amount (YES), the above server system (1) does not stop at simply generating an answer saying "failed to achieve," but can analyze why it was not achieved and perform an improvement schedule data generation step (S13) to resolve it.

[0136] FIG. 7 illustrates the process of identifying production-degrading equipment and deriving improvement schedule data in the improvement schedule generation step according to one embodiment of the present invention.

[0137] The improvement schedule generation step according to one embodiment of the present invention may further include: a step of identifying a production-decreasing facility that, for each of a plurality of facilities related to the schedule data, has produced less than a predetermined threshold value than the production quantity that the facility should have produced according to the schedule data when the actual production quantity is less than the target production quantity; a step of inputting a prompt containing facility data for the production-decreasing facility into a large language model to derive the cause of the production-decreasing facility; and a step of inputting a prompt containing the schedule data, related facility data for each of all facilities related to the schedule data including the production-decreasing facility, and the cause of the production-decreasing facility into a large language model to derive the improvement schedule data.

[0139] As previously explained in FIG. 6, when comparing the actual production amount calculated in the operation judgment step (S12) with the target production amount, if it is determined that the actual production amount falls short of the target production amount, the improvement schedule generation step (S13) may perform a process of identifying which equipment has a decrease in production amount. The equipment in question may be equipment that fails to keep up with the planned speed or yield.

[0140] Specifically, in the improvement schedule generation step (S13), based on the equipment data that can be received from the database (15), the expected production volume and actual performance for each piece of equipment are compared, and equipment with a production decline corresponding to equipment where the difference falls below a preset threshold can be identified. In the example of FIG. 7 (a), an example is illustrated in which equipment B, which is the equipment with a production decline, is identified in the schedule data "Equipment A → Equipment B → Equipment C → Product X production" through such a process.

[0141] In the above improvement schedule generation step (S13), the cause of why equipment B failed to achieve the preset production volume can be identified, and improvement schedule data can be derived based on the cause.

[0143] Figure 7(b) illustrates the process of deriving data on the causes of production decline and improvement schedules through a large language model.

[0144] As illustrated in FIG. 7(b), when the production reduction equipment is identified, equipment data of the production reduction equipment can be collected from the database. The equipment data may preferably include vibration sensor values, motor load, temperature changes, etc.

[0145] In the above improvement schedule generation step (S13), a prompt including one or more of the equipment data of the equipment with reduced production and the related equipment data is input into a large language model to derive the cause of reduced production corresponding to the reason why the equipment with reduced production fails to achieve the expected production volume. For example, causes such as “the tool wear of equipment B reaches a critical value and the processing speed is automatically decelerated” or “the overheating prevention mode is activated due to a lack of cooling water” may be derived as the cause of reduced production.

[0146] Next, in the improvement schedule generation step (S13), the derived cause of production decline, the schedule data, and related equipment data can be input back into a large language model to derive the improvement schedule data. Based on the input data, the large language model can derive alternatives to resolve or bypass the cause of production decline.

[0147] For example, as shown in the example of (c) in Fig. 7, improvement schedule data can be derived, such as reducing the load of equipment B and distributing the remaining volume to other idle equipment D, or urgently scheduling the tool change time of equipment B.

[0149] According to one embodiment of the present invention, when a request is made to verify production volume, the target production volume and the actual production volume are compared and analyzed. If the target is not met, the equipment causing the production decline is identified, and the specific cause is presented. This enables the early elimination of factors causing reduced productivity and the improvement of process yield. Furthermore, by going beyond simple numerical comparison and automatically connecting to an improvement process when production falls short, the delay time between problem recognition and resolution can be minimized.

[0151] FIG. 8 illustrates the process of deriving the final improvement schedule data among a plurality of improvement schedule data generated by reflecting user requirements in the improvement schedule generation step according to an embodiment of the present invention.

[0152] According to one embodiment of the present invention, the user request includes requirements excluding whether normal operation and target production volume are achieved according to the schedule data, and the improvement schedule generation step may further include: a step of inputting the user request, the schedule data, and the related equipment data into a large language model to derive a plurality of improvement schedule data that reflect the requirements included in the user request; and a step of calculating an estimated production volume and an estimated process time for each of the plurality of improvement schedule data, calculating a priority according to a pre-set rule based on the estimated production volume and the estimated process time, and deriving the improvement schedule data with the highest priority among the plurality of improvement schedule data as the final improvement schedule data.

[0154] As illustrated in FIG. 8(a), the user may not simply ask about the current status, but may request specific constraints or simulation conditions, such as, “Equipment C is broken, so please create a schedule excluding Equipment C.” In other words, the user may input a user request that includes specific requirements, which is different from verifying whether the equipment is operating normally or whether a goal has been achieved, as shown in the examples illustrated in FIG. 3 to 6. In this case, the improvement schedule generation step (S13) may perform a process of generating multiple improvement schedule data and deriving the most appropriate one among them.

[0156] As illustrated in FIG. 8(b), when a user request is received in the request reception step (S10) that includes requirements excluding whether normal operation and whether target production volume is achieved, such as the example in FIG. 8(a), in the improvement schedule generation step (S13), the user request, schedule data collected based on the user request, and related equipment data collected based on the schedule data can be input into a large language model to derive multiple improvement schedule data.

[0157] Afterwards, the above server system (1) can input each of the derived multiple improvement schedule data into a large language model or perform a simulation to calculate the expected production volume and expected process time.

[0158] In this context, calculation through simulation may mean inputting variables such as process sequence, operating speed per equipment, and travel time included in each improvement schedule data candidate into a model that generates a virtual environment to perform simulated operation, and deriving quantitative figures for the final product quantity (estimated production volume) and the time required to complete the entire process (estimated process time) when each improvement schedule data is in operation.

[0160] Subsequently, the priority of each candidate can be set according to pre-established rules, such as priority of production volume and priority of delivery deadline compliance. Then, based on the priority, the final improvement schedule data to update the APS (20) can be selected by transmitting it to the schedule update step (S14).

[0162] For example, as shown in Fig. 8(c), when improvement schedule data 1 is predicted to have an 'expected production volume of 500 units and an expected process time of 5 hours', improvement schedule data 2 is predicted to have an 'expected production volume of 400 units and an expected process time of 3 hours', and improvement schedule data 3 is predicted to have an 'expected production volume of 600 units and an expected process time of 8 hours', if the pre-set rule prioritizes 'maximizing production volume', improvement schedule data 3 will be selected, and if 'time efficiency' is important, improvement schedule data 2 or 1 may be selected.

[0164] Additionally, the method may further include the step of calculating a priority based on the user's preference for the priority included in the user request, based on the above-mentioned estimated production volume and the above-mentioned estimated process time, and deriving the improvement schedule data with the highest priority among the plurality of improvement schedule data as the final improvement schedule data.

[0165] Furthermore, if the above user request does not include a preference for the user's priority, the method may further include the step of inputting a prompt including the above user request, the above related equipment data, the above schedule data, and the above multiple improvement schedule data into a large language model to calculate a priority based on a weighting of importance according to the current manufacturing system (2) situation, and deriving the final improvement schedule data with the highest priority among the above multiple improvement schedule data.

[0167] For example, if a user requests, “I am in a hurry for delivery, so please tell me how to finish as quickly as possible,” the server system (1) may determine time efficiency as the highest priority value and give high priority to the candidate with the shortest estimated process time (improvement schedule data 2 in the above example). On the other hand, if a user requests, “It is okay if it takes some time, but we need to produce a lot without defects,” the server system (1) may determine production volume and stability as the highest priority value and give high priority to the candidate with the highest estimated production volume (improvement schedule data 3 in the above example).

[0169] If there is no preference for an explicit priority in the user request, the server system (1) may first input a prompt containing the user request, the schedule data, the related equipment data, and the plurality of improvement schedule data into a large language model. Through this, the importance weight can be calculated by comprehensively considering the current manufacturing process situation (e.g., severity of the cause of failure, remaining delivery date, etc.), and the priority can be determined independently accordingly.

[0170] For example, if the cause of the breakdown is minor and can be resolved by simply replacing parts, a schedule can be selected to operate without interruption by prioritizing 'maintaining production volume.' Alternatively, if the cause of the breakdown is severe and there is a concern that excessive operation could lead to permanent damage, a schedule can be selected to drastically reduce the load or shut down the equipment by prioritizing 'equipment stability.'

[0172] To summarize, the present invention is performed according to the step-by-step flow shown in FIG. 2, but different processing logic and data may be used depending on the type of user request.

[0173] In common, the present invention receives a natural request in the request reception step (S10), collects schedule data and related equipment data in the data collection step (S11), and finally updates the APS (20) with verified improvement schedule data in the schedule update step (S14). However, there may be differences in the type-specific operation judgment step (S12) to the improvement schedule generation step (S13).

[0175] First, in the case of a request asking whether normal operation is being performed as illustrated in FIGS. 3 and 4, the operation judgment step (S12) can identify non-operating equipment by comparing schedule data and related equipment data. In the subsequent improvement schedule generation step (S13), the equipment data of the identified non-operating equipment is analyzed to derive the cause of non-operation, and improvement schedule data can be generated based on this.

[0177] In the case of a request asking whether production volume has been achieved as illustrated in FIGS. 6 and 7, the operation judgment step (S12) may perform an operation to calculate and compare the target production volume based on schedule data and the actual production volume based on historical data included in the equipment data. If the actual production volume falls short of the target production volume, the improvement schedule generation step (S13) may identify the equipment with reduced production efficiency based on the relevant equipment data and derive the cause of the production reduction based on the equipment data of the equipment with reduced production. Then, improvement schedule data may be generated based on the cause of the production reduction.

[0179] Finally, in the case of a request excluding normal operation and production achievement as illustrated in FIG. 8, the operation judgment step (S12) may identify whether there is a conflict between the schedule data and related equipment data and the user request, or transmit the data as is to the next step without a separate judgment. In the improvement schedule generation step (S13), multiple improvement schedule data that satisfy the requirements included in the user request are generated based on the user request, schedule data, and related equipment data, and the most prioritized improvement schedule data among the multiple improvement schedule data can be derived according to a pre-set rule or the user request.

[0181] In this way, by accurately identifying the intent and type of user requests and following an optimized improvement process, it is possible to go beyond simple monitoring and achieve the effect of enabling flexible and in-depth responses to various issues that may occur in the manufacturing site, such as equipment failure, performance degradation, and changes in working conditions.

[0182] In addition, by presenting customized solutions tailored to the situation, it is possible to simultaneously maximize the stability and productivity of process operations.

[0184] Below, the configuration of the database (15) or the process of constructing the database (15), and the process of extracting related equipment data from the database (15) based on the schedule data in the data collection step (S11) are described.

[0186] FIG. 9 illustrates the configuration of a database according to one embodiment of the present invention.

[0187] Specifically, (a) of FIG. 9 illustrates a data storage structure within a database, and (b) illustrates an example of equipment data for a plurality of equipment received from the MES (21) and a definition document describing the same.

[0189] According to one embodiment of the present invention, the server system stores data in the form of a table including a plurality of rows and columns, and defines a mapping rule that defines which column of the equipment data for each of the plurality of equipment corresponds to which column of the integrated data, and the database inputs the data values ​​for each column of the equipment data for each of the plurality of equipment into the corresponding column of the integrated data according to the mapping rule, and the integrated data created or updated by integrating the equipment data for each of the plurality of equipment into one can be stored.

[0191] As illustrated in FIG. 9(a), the database (15) of the present invention can systematically manage and store data collected from a plurality of facilities received from the MES (21) and facility monitoring system (not shown) included in the manufacturing system (2).

[0192] The equipment data receiving unit of the above database (15) can receive equipment data generated in real time from multiple equipment at the site through the above MES (21) and equipment monitoring system.

[0193] At this time, the MES (21) may be a single system that manages data for all of the multiple facilities included in the manufacturing system (2), but may exist as multiple systems separated by process line or type of facility. That is, the database (15) may receive data from one or more MES (21) and facility monitoring systems.

[0194] The received equipment data and definition documents explaining the structure and meaning of the equipment data can be stored in a backup database to prepare for the possibility of data loss in the event of an emergency.

[0196] The integrated data DB management unit of the above database (15) converts equipment data for multiple different equipment into a single standardized integrated data according to pre-set mapping rules and stores and manages it in the main DB. In addition, definition documents explaining the structure and meaning of each equipment data can also be managed.

[0197] In addition, the main DB can store data in a table format containing multiple rows and columns to systematically manage data.

[0199] As illustrated in Fig. 9(b), since multiple equipment (A, B, C) may be of different manufacturers or models, the equipment data (A, B, C) generated by each equipment may have different protocols and data structures.

[0200] As such, equipment data received from different facilities has different data formats and is not standardized, making it difficult to maintain compatibility and integrity during the analysis process.

[0201] For example, the data for Facility A may have columns organized in the order of [Name, Time, Voltage, Current, Location], whereas the data for Facility B may be organized in the order of [Name, Location, Acceleration, Voltage, Current, Time]. Additionally, even if the same physical quantity is represented, the units or notation may differ.

[0203] Meanwhile, the above equipment data may store a definition document in text form that explains information regarding which data type data value each of the multiple columns of the equipment data stores.

[0204] For example, a definition document is stored for each piece of equipment, such as "Column 1 of equipment data A is name, Column 2 is time...", so that a large language model can accurately understand the meaning of the data later.

[0205] The above definition document is a document intended to clearly provide the structure and meaning of equipment data for each facility to the user or a large language model, and may be prepared based on the manual, guide, etc. of the relevant facility.

[0207] According to one embodiment of the present invention, by normalizing various data from heterogeneous equipment according to mapping rules and managing them as integrated data, it is possible to achieve the effect of building a scalable system regardless of the type of equipment. In addition, by preserving the original data in a backup DB and clarifying the meaning of the data through definition documents, a foundation can be established to respond flexibly when new analysis requirements arise in the future.

[0209] FIG. 10 illustrates the configuration of integrated data stored in a database according to one embodiment of the present invention.

[0210] Specifically, FIG. 10(a) illustrates a method for updating equipment data for multiple facilities into integrated data according to mapping rules, and FIG. 10(b) illustrates an example of integrated data generated through the process illustrated in (a).

[0212] The above integrated data DB management unit can input multiple column-specific data values ​​included in each of the multiple equipment data into the data area of ​​the integrated data corresponding to the equipment and the corresponding column, based on a mapping rule that defines the data area in the integrated data corresponding to the column-specific data values ​​for each of the equipment data for multiple equipment.

[0213] Specifically, the above mapping rule may define information regarding which column of the integrated data corresponds to each of the multiple columns of the equipment data for each of the multiple equipment.

[0215] In the example of FIG. 10 (a), the mapping rule for equipment A can be defined to 'correspond column 3 (voltage) of equipment data A to column 1 of integrated data' and 'correspond column 2 (time) of equipment data A to column 3 of integrated data'.

[0216] Similarly, the mapping rule for facility B can be defined to 'match column 4 (voltage) of facility data B to column 4 of integrated data'.

[0218] In this way, the integrated data DB management unit can create and update integrated data by inputting data values ​​for specific columns of equipment data into specific columns of integrated data that correspond to the corresponding columns of equipment data among the multiple columns constituting the integrated data, based on mapping rules stored for each of the multiple equipment or multiple equipment data.

[0219] Integrated data is data generated by aligning and mapping equipment data collected from multiple different facilities according to mapping rules; it is not merely the physical aggregation of individual equipment data, but rather corresponds to a data structure that standardizes the meaning of each piece of equipment data according to consistent attribute definitions.

[0220] In other words, even if data exhibiting the same attributes is stored in different columns or formats across heterogeneous equipment, semantic and formal inconsistencies between equipment data can be resolved and consistent data representation enabled by mapping them to specific columns of a single integrated dataset based on mapping rules. Consequently, the integrated dataset can be used as foundational data for data analysis, comparison, and statistical processing regarding multiple pieces of equipment.

[0222] Meanwhile, in one embodiment of the present invention, data values ​​for all columns of the equipment data are not stored collectively in the integrated data; instead, only specific columns deemed necessary for analysis or the like may be selectively stored in the integrated data. Accordingly, the mapping rule may be defined so that there is no corresponding column in the integrated data for a specific column of the equipment data.

[0224] In one embodiment of the present invention, the mapping rule may be preset by an expert who has a good understanding of the data structure and semantics of the manufacturing system. Specifically, a mapping rule for each of the multiple manufacturing systems may be set by a manager who possesses skilled knowledge regarding the operating principles of the hardware and software, data generation methods, and the meaning and units of data columns of all or part of the multiple manufacturing systems.

[0226] The integrated data generated through a process such as the example in Fig. 10 (a) can be entered such that, as in the example in Fig. 10 (b), the voltage values ​​collected from facility A are aligned in the 'A voltage' column (column 1) of the integrated data, and the voltage values ​​collected from facility B are aligned in the 'B voltage' column (column 4).

[0227] At this time, the integrated data DB management unit may not simply physically combine data, but may generate a definition document for the integrated data that clearly defines which attribute of which equipment each column represents in conjunction with the definition document. For example, Column 1 of the integrated data may be defined as 'voltage of Equipment A' and Column 2 as 'current of Equipment A'. Accordingly, the integrated data can ensure not only formal uniformity but also semantic uniformity of the data.

[0229] FIG. 11 illustrates a process of deriving detailed integrated data containing only data related to schedule data in a data collection step according to an embodiment of the present invention.

[0230] According to one embodiment of the present invention, the data collection step inputs integrated data in which facility data for each of the plurality of facilities is managed integrally and the schedule data into a large language model, derives a related area in the integrated data in which data related to the schedule data is stored, and extracts only the data corresponding to the related area from the integrated data to receive detailed integrated data.

[0232] As illustrated in Fig. 11 (a), the data collection unit (11) can derive a relevant area by inputting a prompt containing a definition document for the schedule data received from the APS (20) and the integrated data into a large language model based on the user request.

[0233] Specifically, the related area may correspond to a data area (column) that must be referenced in the integrated data to derive related facility data for multiple facilities related to the schedule data, as determined by a large language model.

[0234] Alternatively, the related area may correspond to a data area (column) in the integrated data where information related to schedule data is expected to be stored, as determined by a large language model.

[0236] For example, for schedule data such as "A→B→X", a large language model can derive data areas (columns) for Facility A and Facility B as related areas. In this case, the related areas may correspond to columns 1 through 5 of the integrated data where information related to Facilities A and B is stored.

[0237] In other words, columns 6 and 7 of the integrated data, which contain information related to Facility C that is not related to the question, may not correspond to the question-related area.

[0239] When a relevant area is derived through such a process, the data collection unit (11) can generate detailed integrated data by selectively extracting only the data corresponding to the relevant area from the entire integrated data. The detailed integrated data extracted in this way corresponds to relevant equipment data, and subsequently, in the operation judgment step (S12), the large language model can receive the relevant equipment data (preferably the detailed integrated data and the definition document therefor) and schedule data as input and perform an operation judgment.

[0240] For example, as shown in the example of (b) of FIG. 11, the data collection unit (11) can derive detailed integrated data, i.e. related equipment data, from the database (15) by extracting the part describing each of the columns 1 to 5 of the integrated data corresponding to the relevant area from the definition document describing each of the columns 1 to 7 of the integrated data.

[0242] FIG. 12 illustrates the process of updating integrated data according to one embodiment of the present invention.

[0243] Specifically, FIG. 12 (a) illustrates the process of identifying data from the original equipment data when data related to a user request is missing from the integrated data, and FIG. 12 (b) illustrates the process of adding the identified data to the integrated data to update it.

[0245] In one embodiment of the present invention, regarding a new type of user request or an unexpected request to a manufacturing system, instead of outputting a result indicating that an answer to the question cannot be derived because a relevant area is not derived, an answer to the new type of question can be derived through a process of re-searching the data area (reference column) related to the user request in the backed-up equipment data and updating the reference column in the integrated data.

[0247] As shown in the example of FIG. 12 (a), in a situation where a new type of user request (such as "make a schedule to replace the equipment with the highest wear level", which cannot extract schedule data based solely on the user request) is entered, when the data collection unit (11) queries the integrated data (preferably a definition document for the integrated data) of the database (15), the 'wear level' item is not included in the existing mapping rules, so the corresponding information may not exist in the integrated data.

[0248] In this manner, if it is determined that a data value for a specific data type (e.g., wear) of specific equipment data is unnecessary for analysis, and a mapping rule corresponding to a specific column in the integrated data is not set for the column where the data value for that data type of equipment is entered, and thus the corresponding data does not exist in the integrated data, but it is determined that the corresponding data is needed to derive data for a user request, the integrated data DB management unit can perform a process of updating the missing corresponding data in the integrated data.

[0250] Consequently, according to one embodiment of the present invention, in order to derive a response to a new type of user request, data values ​​in a specific column of specific equipment data that were not previously stored in the integrated data are automatically updated in the integrated data, thereby ensuring adaptability capable of responding to various types of user requests.

[0251] In other words, the integrated data (preferably the columns of the integrated data) is not statically fixed but can have a dynamic structure that is continuously added in response to new types of user requests.

[0253] In cases where a relevant area is not derived by the large language model for an input user request, that is, when information is derived that there is no data related to the user request in the integrated data, the integrated data DB management unit inputs the manufacturing system data and a definition document explaining the structure of the manufacturing system data into the large language model to derive information about the reference columns that must be referenced from the equipment data to derive the data related to the user request.

[0255] As described above, for each of the plurality of equipment data, an equipment data definition document describing the structure of the corresponding equipment data is matched and stored, and the data collection unit (11) can input a prompt containing the plurality of equipment data and the definition document matched and stored for each of the plurality of equipment data into a large language model.

[0257] Meanwhile, a large language model that receives such a prompt can derive information on which equipment data among multiple equipment data should be referenced, and which column among multiple columns of said equipment data should be referenced.

[0258] In the above example, the large language model can derive information that it needs to refer to column 6 of the equipment data of equipment A, where the data value for the wear of equipment A is stored, column 5 of the equipment data of equipment B, where the data value for the wear of equipment A is stored, and column 5 of the equipment data of equipment C, where the data value for the wear of equipment A is stored, among equipment data A, B, and C.

[0259] In this way, a specific column of specific manufacturing system data that needs to be referenced can be defined as a reference column. In the example above, the reference columns may correspond to Column 6 of Equipment A, Column 5 of Equipment B, and Column 5 of Equipment C.

[0261] As illustrated in Fig. 12 (b), the integrated data DB management unit can update the data value of the reference column in the integrated data.

[0262] Specifically, as described above, each of the multiple equipment data received from each of the multiple equipment may be backed up and stored separately from the integrated data, and the integrated data DB management unit may update the data value corresponding to the reference column in the backed-up equipment data in the integrated data.

[0263] For example, as shown in the example of Figure 12 (a), when column 6 of equipment A is derived as a reference column, column 8 can be newly created in the integrated data, and the data value of column 6 of the equipment data for equipment A that is backed up can be entered into the newly created column 8 of the integrated data to update the integrated data.

[0264] Likewise, column 5 of equipment B and column 5 of equipment C can be newly created as column 9 and column 10 in the integrated data, and the data values ​​of each column 5 of the equipment data for equipment B and C that are backed up can be entered into the newly created column 9 and column 10 of the integrated data to update the integrated data.

[0266] In addition, as data values ​​for reference columns are entered into the integrated data and updated, the integrated data DB management unit can update the definition document by adding information about the reference columns.

[0267] Specifically, the integrated data DB management department can update the definition document by adding new information about the reference column to the definition document prior to the update, where information about columns 1 through 7 was stored.

[0268] For example, information that column 8 of the integrated data corresponds to the data value for the wear level of Equipment A can be added in text form to the updated definition document.

[0270] When the integrated data is updated in this way, the data collection unit (11) can redefine the area related to the user request (including the wear column) from the updated integrated data as the relevant area, extract the detailed integrated data including it, and provide it to the large language model.

[0272] According to one embodiment of the present invention, even if the data was not defined at the time of initial construction, necessary information is found from the original data in accordance with specific requests from users thereafter, thereby dynamically expanding the integrated data and enabling the implementation of a manufacturing process improvement system that is continuously intelligent and flexibly responds to unexpected analysis requirements.

[0274] FIG. 13 illustrates the configuration of a sub-unit included in an operation judgment unit according to one embodiment of the present invention.

[0275] According to one embodiment of the present invention, the operation determination step is performed in an operation determination unit included in the server system, and the operation determination unit may include one or more of: a process monitoring unit that monitors the real-time process status based on sensing data included in the equipment data; and a state abnormality cause derivation unit that derives the cause corresponding to whether there is a mechanical defect, tool wear, or abnormality in process conditions that caused the state abnormality for the equipment where the state abnormality occurred.

[0277] As illustrated in FIG. 13, the operation judgment unit (12) may be composed of detailed functional modules to deeply analyze the state and cause of a specific process.

[0278] The process monitoring unit can perform the role of comparing and analyzing planned processes with actual processes in progress by linking with MES and CNC (Computer Numerical Control). Specifically, it collects real-time process data such as cutting speed and feed rate, and based on this, can generate basic data for operation judgment information by monitoring in real-time whether the current process is proceeding normally or if a specific event has occurred.

[0279] The abnormal state cause derivation unit can perform logical inference by synthesizing abnormal signs detected by the process monitoring unit or data acquired from external sources. In particular, rather than merely listing phenomena, it can derive fundamental causes such as mechanical defects, tool wear, or abnormal process conditions by analyzing the correlation between abnormal phenomena and potential causes using technologies such as Retrieval-Augmented Generation (GraphRAG).

[0281] FIG. 14 schematically illustrates the internal configuration of a computing device according to one embodiment of the present invention.

[0282] The server system illustrated in FIG. 1 described above may include the components of the computing device (11000) illustrated in FIG. 14.

[0283] As illustrated in FIG. 14, the computing device (11000) may include at least one processor (11100), memory (11200), peripheral interface (11300), input / output subsystem (I / O subsystem) (11400), power circuit (11500), and communication circuit (11600). In this case, the computing device (11000) may correspond to the server system illustrated in FIG. 1.

[0284] The memory (11200) may include, for example, high-speed random access memory, a magnetic disk, SRAM, DRAM, ROM, flash memory, or non-volatile memory. The memory (11200) may include software modules, instruction sets, or various other data required for the operation of the computing device (11000).

[0285] At this time, access to memory (11200) from other components, such as the processor (11100) or peripheral device interface (11300), can be controlled by the processor (11100).

[0286] The peripheral device interface (11300) can connect input and / or output peripheral devices of the computing device (11000) to the processor (11100) and memory (11200). The processor (11100) can perform various functions for the computing device (11000) and process data by executing software modules or instruction sets stored in the memory (11200).

[0287] The input / output subsystem can connect various input / output peripherals to the peripheral interface (11300). For example, the input / output subsystem may include a controller for connecting peripherals such as a monitor, keyboard, mouse, printer, or, if necessary, a touchscreen or sensor to the peripheral interface (11300). According to another aspect, input / output peripherals may be connected to the peripheral interface (11300) without passing through the input / output subsystem.

[0288] The power circuit (11500) can supply power to all or part of the components of the terminal. For example, the power circuit (11500) may include one or more power sources such as a power management system, a battery or alternating current (AC), a charging system, a power failure detection circuit, a power converter or inverter, a power status indicator, or any other components for power generation, management, and distribution.

[0289] The communication circuit (11600) can enable communication with another computing device using at least one external port.

[0290] Alternatively, as described above, the communication circuit (11600) may enable communication with other computing devices by including an RF circuit and transmitting and receiving an RF signal, also known as an electromagnetic signal.

[0291] The embodiment of FIG. 14 is merely an example of a computing device (11000), and the computing device (11000) may have some components shown in FIG. 14 omitted, additional components not shown in FIG. 14 added, or a configuration or arrangement that combines two or more components. For example, a computing device for a communication terminal in a mobile environment may include a touchscreen or sensors, etc., in addition to the components shown in FIG. 14, and the communication circuit (11600) may include a circuit for RF communication of various communication methods (WiFi, 3G, LTE, Bluetooth, NFC, Zigbee, etc.). The components that can be included in the computing device (11000) may be implemented as hardware, software, or a combination of both hardware and software, including one or more integrated circuits specialized for signal processing or applications.

[0292] Methods according to embodiments of the present invention may be implemented in the form of program instructions that can be executed through various computing devices and recorded on a computer-readable medium. In particular, the program according to the present embodiment may be configured as a PC-based program or an application dedicated to a mobile terminal. An application to which the present invention is applied may be installed on a computing device (11000) through a file provided by a file distribution system. For example, the file distribution system may include a file transmission unit (not shown) that transmits the file in response to a request from the computing device (11000).

[0294] The device described above may be implemented as a hardware component, a software component, and / or a combination of a hardware component and a software component. For example, the device and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0295] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computing devices and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0296] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0298] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents. Therefore, other implementations, other embodiments, and equivalents to the claims below are also within the scope of the claims.

Claims

Claim 1 A communicative manufacturing process improvement method for a large language model-based manufacturing system, performed on a server system comprising one or more processors and one or more memories, wherein the server system includes in a database equipment data comprising history data and sensing data for each of a plurality of equipment included in the manufacturing system, received from a Manufacturing Execution System (MES) and an equipment monitoring system for the manufacturing system, and the manufacturing process improvement method comprises: a request reception step of receiving a user request in natural language form from a user terminal; a data collection step of receiving schedule data related to the user request from an Advanced Planning and Scheduling (APS) for the manufacturing system and collecting related equipment data including equipment data for each of all equipment related to the schedule data from the database; and an operation judgment step of inputting the user request, the schedule data, and the related equipment data into an internal or external large language model to derive operation judgment information for each of a plurality of equipment related to the schedule data, including whether the equipment has normally performed the operation that the equipment is required to perform according to the schedule data. A method for improving a manufacturing process through communication, comprising: an improvement schedule generation step in which, when equipment that has not normally performed an operation corresponding to the schedule data is identified in the operation judgment information, the improvement schedule data is re-established based on equipment data for the equipment, the schedule data, and the user request, so that the equipment achieves a target production quantity or target completion time according to the schedule data; and a schedule update step in which the improvement schedule data is transmitted to the APS to cause the manufacturing system to perform process operations according to the improvement schedule data. Claim 2 The method for improving a communicative manufacturing process according to claim 1, wherein the improvement schedule generation step comprises: a step of deriving the cause of non-operation of the non-operation equipment by inputting a prompt containing equipment data for the non-operation equipment into a large language model when a non-operation equipment corresponding to equipment that has not normally performed an operation corresponding to the schedule data is identified in the operation judgment information; and a step of deriving the improvement schedule data by inputting the relevant equipment data containing equipment data for each of all equipment related to the schedule data including the non-operation equipment, and the prompt containing the cause of non-operation into a large language model. Claim 3 A communicative manufacturing process improvement method according to claim 1, wherein the improvement schedule generation step calculates an expected production volume when the manufacturing system operates according to the improvement schedule data, verifies the validity of the improvement schedule data based on whether the expected production volume is greater than or equal to the existing production volume according to the schedule data, and the schedule update step transmits the improvement schedule data to the APS only when the validity of the improvement schedule data has been verified. Claim 4 A method for improving a manufacturing process using communication, wherein, in claim 1, the user request includes a production quantity verification request inquiring whether the actual production quantity produced by the manufacturing system according to the schedule data has achieved the target production quantity that the manufacturing system should have produced when following the schedule data, and the operation judgment step includes: a step of calculating the target production quantity according to the schedule data and receiving history data for each of a plurality of facilities related to the schedule data to calculate the actual production quantity when the user request corresponds to a production quantity verification request; and a step of comparing the target production quantity with the actual production quantity to determine whether the target production quantity has been achieved. Claim 5 In claim 4, the improvement schedule generation step further comprises: a step of identifying a production-decreasing facility that, for each of a plurality of facilities related to the schedule data, has produced less than a preset threshold value than the production quantity that the facility should have produced according to the schedule data when the actual production quantity is less than the target production quantity; a step of inputting a prompt containing facility data for the production-decreasing facility into a large language model to derive the cause of the production-decreasing facility; and a step of inputting a prompt containing the schedule data, relevant facility data for each of all facilities related to the schedule data including the production-decreasing facility, and the cause of the production-decreasing facility into a large language model to derive the improvement schedule data; a communicative manufacturing process improvement method. Claim 6 A method for improving a manufacturing process through communication according to claim 1, wherein the server system stores data in the form of a table including multiple rows and columns, and defines a mapping rule that defines which column of the equipment data for each of the multiple equipment corresponds to which column of the integrated data, and the database stores integrated data in which the equipment data for each of the multiple equipment is integrated into one and created or updated. Claim 7 A method for improving a communicative manufacturing process according to claim 1, wherein the data collection step inputs integrated data in which equipment data for each of the plurality of equipment is managed integrally and the schedule data into a large language model, derives a related area in the integrated data in which data related to the schedule data is stored, and extracts only the data corresponding to the related area from the integrated data to receive detailed integrated data. Claim 8 A method for improving a communicative manufacturing process according to claim 1, wherein the operation determination step is performed in an operation determination unit included in the server system, and the operation determination unit comprises one or more of: a process monitoring unit that monitors a real-time process state based on sensing data included in the equipment data; and a state abnormality cause derivation unit that derives a cause corresponding to whether there is a mechanical defect, tool wear, or abnormality in process conditions that caused the state abnormality for the equipment where the state abnormality occurred. Claim 9 A server system comprising one or more processors and one or more memories, and performing a communicative manufacturing process improvement method for a large language model-based manufacturing system, wherein the server system stores equipment data including history data and sensing data for each of a plurality of equipment included in the manufacturing system, received from a Manufacturing Execution System (MES) and an equipment monitoring system for the manufacturing system; a request receiving unit that receives a user request in natural language form from a user terminal; a data collection unit that receives schedule data related to the user request from an Advanced Planning and Scheduling (APS) for the manufacturing system and collects related equipment data related to the schedule data from the database; an operation judgment unit that inputs the user request, the schedule data, and the related equipment data into an internal or external large language model to derive operation judgment information for each of a plurality of equipment related to the schedule data, including whether the equipment has normally performed the operation that it is required to perform according to the schedule data; and when equipment that has not normally performed the operation corresponding to the schedule data is identified in the operation judgment information, the A server system comprising: an improvement schedule generation unit that generates improvement schedule data re-established based on equipment data for equipment, the schedule data, and the user request, so that the equipment achieves a target production volume or target completion time according to the schedule data; and a schedule update unit that transmits the improvement schedule data to the APS to cause the manufacturing system to perform process operations according to the improvement schedule data.

Images