Method for optimizing process using generative ai and system used therefor

Abstract

The present invention relates to a process optimization method using generative AI and a system therefor, and more specifically, to a process optimization method and system capable of searching or reasoning causal relationships between a plurality of steps and between variables accompanying an arbitrary process by an artificial intelligence algorithm and deriving corresponding hypotheses, and optimizing a process using the same. And simulation and verification are carried out.

Description

TECHNICAL FIELD

[0001] The present invention relates to a process optimization method and system using generative AI, and more particularly, to a process optimization method and system capable of retrieving or reasoning about the causal relationship between a plurality of steps accompanying an arbitrary process and between variables by an artificial intelligence algorithm, and deriving a corresponding hypothesis to perform simulation and verification. BACKGROUND

[0002] Generally, a process of producing a certain product is complex and includes various steps, and is characterized in that there are very many variables interacting with each other in the process. Therefore, although there is a demand to optimize the process, efficient optimization is almost impossible through conventional methods due to the complexity of the process.

[0003] On the other hand, in recent years, deep learning technology has shown excellent performance in learning and predicting complex patterns based on a large amount of data. In particular, generative AI has shown full availability in the field of process optimization by its ability to generate new data or change existing data.

[0004] The present invention is proposed in consideration of the demand for optimization of the process and the applicability of the generative AI algorithm, and aims to improve the process optimization process that has had technical limitations in the past by using an artificial intelligence algorithm capable of causal relationship retrieval and causal relationship reasoning in the entire process.

[0005] Also, the invention to be described through the detailed description can not only eliminate the technical problems introduced above, but also be invented in order to provide additional technical elements that cannot be easily invented by those skilled in the art.

[0006] PRIOR ART DOCUMENT

[0007] PATENT DOCUMENT

[0008] Patent Document 1: Korean Patent Publication No. 10-2023-0089693 (2023.06.21) SUMMARY

[0009] PROBLEMS TO BE SOLVED BY THE INVENTION

[0010] The present invention aims to analyze the causal relationship between variables in a complex production process, thereby efficiently deriving key factors for optimizing the process.

[0011] Also, the present invention aims to generate various process improvement hypotheses supported by logical reasoning, and to realize an environment that can systematically verify the hypotheses.

[0012] Furthermore, the purpose of this invention is to simulate hypotheses derived through causal reasoning to provide an environment in which the anticipated effects applicable to actual processes can be known, thereby improving user convenience.

[0013] Furthermore, its purpose is to enable users to intuitively understand the analysis results and hypotheses through a user-friendly interface, thereby allowing users to make decisions easily.

[0014] Methods for solving problems

[0015] The present invention was made to solve the problems described above. The method for process optimization provided by the present invention, which is a process optimization system including a central processing unit and a memory, may include: (a) a process data receiving step, receiving multiple process data; (b) a causal relationship derivation step, analyzing the process data and deriving causal relationships between variables; (c) a causal relationship inference step, calculating the strength of the causal relationship between the variables with reference to the derived causal relationship; (d) a simulation step, simulating conditions for changing specific variables with reference to the causal relationship inference results calculated in the causal relationship inference step; and (e) a hypothesis generation step, generating at least one hypothesis for process optimization with reference to the simulation results calculated in the simulation step.

[0016] Furthermore, in the method for optimizing the process, the causal relationship derivation step may include the following steps: in order to understand the characteristics or patterns of the process data, retrieving the process data; deriving the correlation between any variables contained in the process data; selecting the main variable among the arbitrary variables; and deriving the causal relationship between the variables based on the main variable.

[0017] Furthermore, in the method for optimizing the aforementioned process, the causal relationship derivation step may include the following steps: inputting the causal relationship derivation result into a generative AI-assisted system, wherein the causal relationship derivation result is generated by executing the causal relationship derivation step; and receiving AI parsing data corresponding to the causal relationship derivation result, wherein the AI ​​parsing data is generated by the generative AI-assisted system.

[0018] Furthermore, in the method for optimizing the process, the causal relationship derivation step may further include a step of analyzing the received process data and the AI ​​parsing data to re-derive the causal relationship between the variables.

[0019] Furthermore, in the method for optimizing the aforementioned process, the AI-analyzed data may include causal relationship descriptions generated with reference to stored literature content and sources related to the literature.

[0020] Furthermore, in the method for optimizing the process, in the causal relationship derivation step, in addition to inputting the process data, at least one of experimental data or prior knowledge data may be input to derive the causal relationship.

[0021] Furthermore, in the method for performing the process optimization, it is possible to link the process optimization as the final indicator of the objective with any variable in the process data during the step of retrieving the process data.

[0022] Furthermore, in the method for optimizing the aforementioned process, the causal reasoning step may include the following steps: inputting the causal reasoning result into a generative AI-assisted system, wherein the causal reasoning result is generated by executing the causal reasoning step; and receiving AI parsing data corresponding to the causal reasoning result, wherein the AI ​​parsing data is generated by the generative AI-assisted system.

[0023] Furthermore, in the method for optimizing the aforementioned process, the causal reasoning step may further include a step of analyzing the received causal relationship derivation results and the AI ​​parsing data to recalculate the strength between the causal relationships.

[0024] Furthermore, in the method for optimizing the process, it is possible to perform simulation by referring to the causal relationships and causal strengths between specific variables included in the causal reasoning results, and by referring to simulation condition information arbitrarily generated by the process optimization system or input by any user, during the simulation step.

[0025] Furthermore, in the method for optimizing the aforementioned process, it is possible that, in the hypothesis generation step,

[0026] Based on the simulation results, hypotheses are generated that include specific causal relationships and the effects of those causal relationships. Literature databases are then consulted to determine whether the hypotheses contain known causal relationships.

[0027] Furthermore, in the method for optimizing the process, the hypothesis generation step may include the following steps: inputting at least one hypothesis into a generative AI-assisted system, wherein the at least one hypothesis is generated by performing the hypothesis generation step; and receiving AI parsing data corresponding to the hypothesis, wherein the AI ​​parsing data is generated by the generative AI-assisted system.

[0028] On the other hand, another embodiment of the present invention provides a method for process optimization by a process optimization system including a central processing unit and a memory, which may include: (a) a process data receiving step, receiving multiple process data; (b) a causal relationship derivation step, analyzing the process data and deriving causal relationships between variables; (c) a causal relationship inference step, calculating the strength of the causal relationship between the variables with reference to the derived causal relationship; (d) a hypothesis generation step, generating at least one hypothesis for process optimization with reference to the causal relationship inference result calculated in the causal relationship inference step; and (e) a simulation step, simulating conditions for changing specific variables with reference to information about the hypothesis.

[0029] Furthermore, in the method for optimizing the process, if multiple hypotheses are generated in the hypothesis generation step, a step of selecting the hypothesis required for simulation may be included after the hypothesis generation step, in which only the selected hypothesis is simulated.

[0030] On the other hand, another embodiment of the present invention provides a process optimization system including a central processing unit and a memory. The central processing unit executes instructions stored in the memory for executing a process optimization method. The method for executing the process optimization method includes: (a) a process data receiving step, receiving multiple process data; (b) a causal relationship derivation step, analyzing the process data and deriving causal relationships between variables; (c) a causal relationship inference step, calculating the strength of the causal relationship between the variables with reference to the derived causal relationship; (d) a simulation step, simulating conditions for changing specific variables with reference to the causal relationship inference result calculated in the causal relationship inference step; and (e) a hypothesis generation step, generating at least one hypothesis for process optimization with reference to the simulation result calculated in the simulation step.

[0031] Furthermore, another embodiment of the present invention provides a process optimization system including a central processing unit and a memory. The central processing unit executes instructions stored in the memory for executing a process optimization method. The method for executing the process optimization method includes: (a) a process data receiving step, receiving multiple process data; (b) a causal relationship derivation step, analyzing the process data and deriving causal relationships between variables; (c) a causal relationship inference step, calculating the strength of the causal relationship between the variables with reference to the derived causal relationship; (d) a hypothesis generation step, generating at least one hypothesis for process optimization with reference to the causal relationship inference result calculated in the causal relationship inference step; and (e) a simulation step, simulating conditions for changing specific variables with reference to information about the hypothesis.

[0032] Invention Effects

[0033] According to the present invention, it has the effect of improving process efficiency and product quality by applying optimal process conditions derived through causal relationship analysis and generative AI.

[0034] Furthermore, according to the present invention, the number of actual experiments is reduced through simulation and verification, thus preventing unnecessary waste of resources.

[0035] Furthermore, according to the present invention, new process improvement variables or hypotheses that have not been previously proposed can become the research object, thereby enabling innovative improvement of process effectiveness.

[0036] Furthermore, according to the present invention, information is provided in a manner that allows users to intuitively understand the matters required to improve the process and the basis thereof, thereby helping users to make decisions easily.

[0037] Furthermore, according to the present invention, by repeatedly reflecting prior knowledge, deriving causal relationships, and providing auxiliary explanations based on generative AI, it is possible to ensure knowledge or information throughout the entire process using various forms of data. Attached Figure Description

[0038] Figure 1 This is a diagram used to conceptually and easily understand the process optimization system provided by this invention.

[0039] Figure 2 The detailed structure of the generative AI-assisted system is shown.

[0040] Figure 3 The process optimization method provided in the first embodiment is shown.

[0041] Figure 4 A graph showing the results of deriving the causal relationship is displayed.

[0042] Figure 5 A graph showing the results of calculating the strength of each causal relationship among the retrieved causal relationships and ultimately inferring the causal relationship is presented.

[0043] Figure 6 The process optimization method provided in the second embodiment is shown.

[0044] Explanation of reference numerals in the attached figures

[0045] 10: Process optimization system; 100: Process optimization calculation system; 200: Generative AI-assisted system. Detailed Implementation

[0046] A more detailed understanding of the objectives, technical structure, and corresponding effects of this invention will become clearer from the following detailed description based on the accompanying drawings. Embodiments of the invention will be described in detail with reference to the drawings.

[0047] The embodiments disclosed in this specification should not be construed as limiting the scope of the invention. For those skilled in the art, the descriptions in this specification, including the embodiments, will clearly have various applications. Therefore, any embodiments described in the detailed description of the invention are merely illustrative and are not intended to limit the scope of the invention to the embodiments.

[0048] The functional block diagrams presented in the accompanying drawings and described below are merely feasible implementation examples. Other functional block diagrams can be used in other implementations without departing from the spirit and scope of the detailed description. Furthermore, while more than one functional block diagram of the present invention is presented individually, one or more functional block diagrams of the present invention may be a combination of various hardware and software structures performing the same function.

[0049] Furthermore, the statement that it includes certain constituent elements is an "open-ended" statement, which merely indicates the existence of the constituent element and should not be interpreted as excluding additional constituent elements.

[0050] Furthermore, when it is mentioned that a constituent element is "connected" or "linked" to another constituent element, it should be understood that it can be directly connected or linked to other constituent elements, but there may also be other constituent elements between the two.

[0051] Figure 1 This diagram is for conceptual understanding of the process optimization system provided by the present invention. Referring to the accompanying drawings, the present invention is characterized in that when process data, such as detailed steps included in the process and multiple variables obtainable for each process, are input into the process optimization system 10, the causal relationships between the variables are analyzed and simulations are performed based on the process data, thereby providing at least one or more hypotheses for optimizing the process. Furthermore, in this process optimization process, generative AI can be actively utilized. Throughout the process optimization calculations, including deriving causal relationships between variables in the process, inferring causal relationships, and anticipating simulation results, this generative AI, referring to literature data related to the process, provides the user with visual information on the basis for deriving or inferring causal relationships or anticipating simulation results, thereby enabling the user to easily understand the variables required for process optimization and the degree of adjustment.

[0052] Referring to the accompanying drawings, it can be confirmed that the process optimization system 10 is designed to include two subordinate systems: a process optimization calculation system 100 and a generative AI-assisted system 200. The process optimization calculation system 100 can be implemented to analyze the causal relationships between variables based on process data and generate hypotheses accordingly. The generative AI-assisted system 200 can be implemented to work in conjunction with the process optimization calculation system 100 to provide the information required for each optimization calculation step. On the other hand, it should be understood that the illustrated method shown in the accompanying drawings, dividing the process optimization system 10 into two subordinate systems, has been chosen in this detailed description for ease of understanding of the invention. However, the process optimization system 10 can be composed of more subordinate systems or implemented by a single system as needed.

[0053] The process optimization calculation system 100 may include a causal relationship derivation unit 110, a causal relationship reasoning unit 120, a simulation unit 130, and a hypothesis generation unit 140. The generative AI-assisted system 200 may include a generative AI implementation unit 210, a data retrieval unit 220, a literature database 230, and a data collection unit 240. Detailed structures of these components will be discussed in [the following sections / sections]. Figure 3 and Figure 2 This will be discussed later in the explanation.

[0054] On the other hand, by utilizing the process optimization system 10 provided by this invention, various processes can be optimized. For example, processes for producing biological products can be listed. Specifically, processes for producing amino acids or nucleic acids include the cultivation and fermentation of various microorganisms under optimal conditions. During this process, process data, including various variables such as pH, temperature, pressure, airflow, and culture medium composition, are managed. In particular, the fermentation process can be divided into Jar, Seed, and Main steps based on the growth stage of the microorganisms, with different environments and corresponding measurable and controllable variables for each step. Furthermore, in each step, various variables can be measured or controlled, such as environmental variables like temperature, pH, and OD (Optical Density), the culture medium to be added as nutrients for the microorganisms (glucose, salt, raw sugar, etc.), the timing and frequency of adding and adjusting the culture medium and environmental factors. These variables can be measured by measuring means installed on the process equipment or directly input into a terminal by the process administrator (operator) for storage. In a process involving such complex steps and variables, the metrics that can be optimized include the production efficiency, output, and time consumed in the production of the final product. Furthermore, the growth time and output measured in each step of the process can also be included as metrics that can be optimized.

[0055] On the other hand, the main feature of the process optimization system provided by this invention is that, in an environment where the causal relationships between variables are complex and the user cannot fully grasp the causal relationships between all variables in the process, it can receive input from variables monitored during product production, and thereby analyze which variables have causal relationships and the strength of these causal relationships, and based on the results, provide the user with at least one hypothesis for optimizing the amino acid production process. As a more specific example, it can be derived that when pH is adjusted in the microbial culture step, the growth rate of microorganisms increases, thereby increasing the yield of amino acids. Furthermore, it can quantitatively predict how much the yield of amino acids will increase when the pH is adjusted to a certain level.

[0056] On the other hand, the process optimization method provided by this invention mentions that generative AI can be actively utilized. In the causal relationship analysis and hypothesis generation steps described above, the generative AI reference data is used to show the user what basis supports the derivation and reasoning of the above-mentioned causal relationship. This allows the user to directly grasp the variables that need to be adjusted in process optimization, or to assist the user (operator) in referring to various hypotheses proposed by the generative AI, and to gain inspiration through more highly developed human thinking.

[0057] Figure 2 The detailed structure of the generative AI-assisted system 200 is shown in the accompanying drawings. Referring to the drawings, the generative AI-assisted system 200 may include a generative AI implementation unit 210, a data retrieval unit 220, a literature database 230, and a data collection unit 240.

[0058] The generative AI implementation unit 210 is a structure that enables generative AI to operate in at least a portion of the steps included in process optimization. This generative AI implementation unit 210 can be designed to generate vectors about arbitrary input data, perform operations based on the desired vectors, and then generate text from the results. In the generative AI implementation unit 210, a pre-trained large language model (PLLM) for processing text data or a model that uses a RAG structure and generates responses using external reference data can be utilized.

[0059] The generative AI implementation unit 210 can be designed to receive input at each step of the process optimization process. This input can be a prompt arbitrarily entered by the user or automatically entered by the process optimization calculation system 100. The content of the prompt will vary depending on the step of the process optimization, but preferably it can include the result value derived in that step and / or the result value derived in the preceding steps. It can also include a request for a literature-based explanation derived from the result value (such as causal relationships, rationality, counterarguments, variables that need to be considered, or the expected effect, etc.). Furthermore, the prompt can include professional knowledge literature (or, content converted into vector form) that the generative AI can refer to when outputting the literature-based explanation, thereby enabling the generative AI to provide an accurate response. On the other hand, the prompt can also contain only information generated autonomously by the generative AI. That is, even without a result value derived in a specific step or a result value derived in a preceding step, by including only the information generated by the generative AI in the prompt, it is possible to guide the generative AI to provide an innovative response.

[0060] The data retrieval unit 220 efficiently retrieves information that matches the queries input by the generative AI. The data retrieval unit 220 can directly connect to the literature database 230 to access the required literature data, and can assist the generative AI in generating appropriate responses by retrieving and obtaining literature data that corresponds to the prompts input at each step of each optimization process. To this end, the data retrieval unit 220 may have the following functions: vectorization of text data; and vectorization of the requests (queries) contained in the prompts, and retrieval of data with similar vector representations from the vector store.

[0061] The literature database 230 is a structure for storing and managing professional knowledge literature related to the process, playing a crucial role in process optimization, especially in assisting generative AI to calculate accurate analytical results. For reference, literature is written in text form, referring to books, documents, etc., containing explanations of arbitrary factual relationships; such literature can be converted into digital format and stored in the literature database 230. The literature database 230 can store process-related research papers, technical reports, research notes, textbooks, etc. The literature stored in the literature database 230 can exist in various forms, such as being in the form of digital files, original text converted to digital format, converted to text data through text recognition technology, or converted to vector format through a language model. For reference, the literature database 230 can also store hypotheses whose causal relationships or correlations have been verified as valid to a certain level or higher by the process optimization system provided by this invention. Specifically, if it is verified that a specific condition calculated by the process optimization system provided by this invention has a meaningful impact on the final product of the test (simulation) results, the hypothesis related to that condition is stored in the literature database 230, thus enabling it to be used as prior knowledge in subsequent process optimization processes.

[0062] Finally, the data collection unit 240 continuously retrieves professional knowledge literature from external sources and stores new professional knowledge literature in the literature database 230 when it is found. The data collection unit 240 can be configured to periodically search for keywords to determine whether there is new professional knowledge literature related to the process, and when such professional knowledge literature is found, the data collection unit 240 can provide at least a portion of the literature content, the source of the literature, the author of the literature, and other relevant information so that users can judge the validity and reliability of the literature.

[0063] The above is for reference only. Figure 1 as well as Figure 2 A conceptual introduction to process optimization systems and methods is provided.

[0064] Before proceeding with the formal explanation, let's first introduce the hardware implementation components of the aforementioned process optimization system 10. It will be understood that the process optimization method provided by this invention can be executed by the process optimization system 10 or a computing device having a central processing unit and memory. Such systems or computing devices can include portable terminals such as smartphones, personal digital assistants (PDAs), and tablet computers, as well as terminals fixed in a specific location such as desktop computers. The central processing unit can also be referred to as a controller, microcontroller, microprocessor, microcomputer, etc. Furthermore, the central processing unit can be implemented through hardware, firmware, software, or a combination thereof. When implemented in hardware, it can be implemented by ASIC (application specific integrated circuit), DSP (digital signal processor), DSPD (digital signal processing device), PLD (programmable logic device), FPGA (field programmable gate array), etc. When implemented in firmware or software, the firmware or software can be configured to include modules, programs, or functions that perform the functions or actions described above. Furthermore, the memory can be implemented using ROM (Read Only Memory), RAM (Random Access Memory), EPROM (Erasable Programmable Read Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, SRAM (Static RAM), HDD (Hard Disk Drive), SSD (Solid State Drive), etc.

[0065] Depending on the circumstances, the system 10 may be a server, which can be a device that stores and executes a set of programs, i.e., instructions, that are actually used to implement the process optimization method provided by the present invention. The server may be in the form of at least one server computer managed by a specific user, or it may be in the form of a cloud server provided by another company, i.e., a cloud server that users can use by registering as a member.

[0066] Furthermore, depending on the circumstances, the process optimization method provided by this invention can also be executed on a cluster system consisting of multiple systems or computing devices, rather than on a single system 10. Within the cluster system, it can be configured so that the multiple computing devices required to execute the process optimization method perform different operations.

[0067] The following section, with reference to the accompanying drawings, describes each step of the process optimization method.

[0068] Figure 3 The process optimization method provided by the first embodiment of the present invention is illustrated step by step. Referring to the accompanying drawings, it can be confirmed that the process optimization method can be executed by the detailed structure of the process optimization calculation system 100, namely the causal relationship derivation unit 110, the causal relationship reasoning unit 120, the simulation unit 130, and the hypothesis generation unit 140. At this time, each detailed structure can be implemented by an algorithm module implemented by a set of instructions.

[0069] The process optimization method provided in the first embodiment may first include step S101, in which the causal relationship derivation unit 110 receives process data. The causal relationship derivation unit 110 may be designed to receive process data by directly communicating with a component that monitors the process system (not shown) (including a sensor, CCTV (Closed Circuit Television), or other component capable of monitoring or monitoring the process status) or with the terminal (e.g., a smartphone, a portable device for process data) of an operator who inputs various process data at the location (factory) where the process is completed, or by receiving process data from a process data database that stores process data.

[0070] Process data can be categorized in various ways, including, for example, temperature (temperature of each process step, temperature of each incubator, etc.), pH, pressure (pressure of the incubator, pressure of the pipeline for transporting the culture medium / microorganisms, etc.), airflow, dissolved oxygen, carbon dioxide concentration, sampling volume, type of culture medium, composition of culture medium, volume of culture medium, culture medium ratio, amount of culture medium added, time point of culture medium addition, culture medium replacement cycle, microbial density that can be monitored during the incubation process, energy consumption of the process, type and amount of waste generated by the process, and temperature and humidity within the plant. Furthermore, in addition to information related to the process itself, as mentioned above, information about the workers within the plant or their work can also be included as a type of process data. That is, process data can also include information related to the materials used in the process, information monitored during the process, information related to the location where the process is completed, and information related to the personnel (workers) involved in the process. Necessary information can be selected for subsequent derivation or inference of causal relationships.

[0071] On the other hand, the process data can be preprocessed before and after step S101. For example, the following processing can be performed: excluding data that is above the baseline value; excluding data that is set as unnecessary according to user settings; and excluding data that exceeds the set normal range. In addition, for missing values, data can be supplemented using known methods such as average, median, and KNN. This data preprocessing can be performed before the process data is passed to the causal relationship derivation unit 110, or it can be performed by the causal relationship derivation unit 110 or by another preprocessing unit (not shown) included in the process optimization system 10.

[0072] After receiving process data and completing data preprocessing, the causal relationship derivation unit 110 can perform calculations (S102) to derive causal relationships between specific variables throughout the entire process from the process data. This step may include processes for retrieving data, deriving correlations between variables, deriving main variables, and deriving causal relationships between specific variables.

[0073] The primary purpose of data retrieval is to analyze previously collected process data to understand its characteristics or patterns. Another objective is to further identify problems that can be discovered from these characteristics or patterns. In this data retrieval process, process data is analyzed in conjunction with a target final indicator (e.g., productivity), allowing for the retrieval of variables influencing the target indicator starting from the first process derived from causal relationships. Furthermore, the data retrieval process can utilize not only statistical calculations but also algorithms for discovering patterns or regularities in recurring data and clustering algorithms to analyze the characteristics or patterns of the data. In another aspect, the data retrieval process can utilize visual means (charts, scatter plots, etc.) to present the distribution, trends, outliers, etc., of the data to provide users with information about the characteristics, patterns, or problems of the process data. This provides an environment where users can intuitively understand the relationships between variables (the relationships between variables monitored in different steps or variables monitored in the same step) or predict problems in the process.

[0074] Following the data retrieval process, the process of deriving the correlation between variables can be performed. If the previous process was about analyzing the characteristics and patterns of process data, then this process can quantitatively determine whether any variable has an impact on other variables (whether there is a correlation) and the degree of impact (the strength of the correlation).

[0075] In the process of deriving correlations, the existence and strength of correlations can be determined through correlation analysis between the target indicator and specific variables, as well as regression analysis. The purpose of correlation analysis is to understand the correlations between various variables using matrices or heatmaps. Therefore, the results of correlation analysis can reveal whether process variables influence each other and the strength of that influence. Furthermore, by providing a visual output to the user, it allows for intuitive understanding. The purpose of regression analysis is to functionalize the relationship between any specific variable and other variables, reflecting the trend after functionalization, to predict the impact of that specific variable on other variables. The results of regression analysis can also reveal whether process variables influence each other and the strength of that influence.

[0076] On the other hand, following the correlation derivation process, the process of deriving the principal variables can be performed. In this process, specific variables or variables analyzed to have a significant impact on the process can be selected. This process may include the following steps: based on the results obtained from the previous analysis of process data, using a machine learning model to predict the target indicator, and understanding which variables contribute to the target indicator and to what extent, to select the principal variables. Furthermore, this process may also include the following steps: embedding variables, and using the embedding of the target indicator and other variables to select variables that have a significant mutual influence.

[0077] After deriving the main variables, the process of deriving the causal relationships between variables in the process can be performed. This is the core operation performed in the causal relationship derivation section 110. This process can be understood as defining a causal relationship based on the relationships between variables obtained in the previous data retrieval and correlation derivation steps, defining which variables affect other variables in the actual process. In this detailed description, the terms "correlation" and "causation" are used in contrast. If correlation indicates the tendency of data obtained by quantitatively understanding the association between variables, then causation indicates the relationship in which any variable causes changes in other variables. Therefore, in addition to reflecting the results of correlation analysis between variables, it is also necessary to reflect the various mechanisms in the process to define a causal relationship.

[0078] The causal relationship derivation unit 110 can be implemented to discover causal relationships between variables in a process using a causal relationship derivation algorithm (or model). As input, previously analyzed process data can be input, as well as experimental data—that is, experimental data conducted separately from the process and recorded in its process. The purpose of inputting experimental data is to enable the causal relationship derivation algorithm to reflect the influence between variables in process steps that cannot be grasped through simple quantitative analysis, by referencing not only data obtained in the process but also data obtained in a strictly controlled experimental environment. This experimental data can be provided from a literature database—a database storing various references—which will be mentioned later in this detailed description, and this experimental data can be continuously updated as various experiments are conducted.

[0079] On the other hand, the data that can be input into the causal relationship derivation algorithm can also include prior knowledge data. For example, by inputting the following three quantified prior knowledge data into the causal relationship derivation algorithm, the aforementioned prior knowledge can be reflected when deriving causal relationships. These three prior knowledge data are: prior knowledge about the influence of the concentration or amount of phosphate on the time consumed in fermentation; prior knowledge about the influence of the type or concentration of minerals on the productivity of the final product; and prior knowledge about the influence of sugar concentration on the fermentation yield. The input of such prior knowledge data can be directly entered by the user through an input interface provided to the user, or it can also be implemented by querying prior knowledge data (literature data) pre-stored in a literature database. When inputting prior knowledge data, the reliability or probability of the prior knowledge data can be input together. For example, the relationship between specific variables can be assigned values ​​of 0, 1, or between 0 and 1 as prior knowledge data. It will be understood that the reliability or probability of prior knowledge data does not necessarily have to be evaluated through specific values; various forms of evaluation media capable of evaluating levels can exist.

[0080] In another aspect, the data that can be input into the causal relationship derivation algorithm can also include AI-analyzed data, which is the analysis result of the causal relationships calculated by generative AI. Generative AI can autonomously analyze the causal relationships between variables by referring to given process data and correlation derivation results. In this process, the analysis results of the causal relationships can be generated as AI-analyzed data. This AI-analyzed data can be generated by the generative AI-assisted system 200 by referring to professional knowledge data in a stored literature database. In this case, AI-analyzed data can be understood as a type of literature-based data.

[0081] On the other hand, the causal relationship derivation unit 110 can receive the AI-analyzed data again and use it as a reference when deriving (or re-deriving) causal relationships, thereby enabling the retrieval of causal relationships that are difficult for humans to intuitively understand, i.e., potential causal relationships. The AI-analyzed data can be generated by performing causal relationship analysis on the generative AI starting from the time point of input process data and experimental data, and can be utilized by inputting the AI-analyzed data starting from the initial step of executing the causal relationship derivation algorithm by the causal relationship derivation unit 110. In another aspect, the causal relationship derivation result obtained by executing the causal relationship derivation algorithm by the causal relationship derivation unit 110 can also be used as input to the causal relationship derivation unit 110 again (S103-1). At this time, it is also possible to store the causal relationship derivation result in a literature database and process it in a way that it can be retrieved by the causal relationship derivation unit 110, or immediately re-input the causal relationship derivation result as input to the causal relationship derivation unit 110, thereby deriving more accurate causal relationships by repeatedly executing the causal relationship derivation algorithm. For example, when a prompt input corresponding to the relationship between the main variables that affect productivity and productivity in the request-generating AI generates a response such as [Causal relationship: Oxygen concentration affects pH change; Basis: Oxygen concentration affects the aerobic and anaerobic metabolism of microorganisms in the fermentation process, which affects the generation of metabolites and pH change], the causal relationship in such a generated response can again become the input to the causal relationship derivation unit 110, thereby helping to derive the causal relationship again.

[0082] Figure 4 An example of a chart visually presenting the results of causal relationships is shown. Referring to the accompanying figure, the causal relationship results can be labeled with causal arrows indicating which variable influences other variables, and the positions of each variable can be configured to represent specific steps in the process. Furthermore, in the chart with the variables configured, the vertical axis can represent a time axis, with variables occurring in the latter part of the process, arranged from top to bottom. Additionally, variables for which no causal relationship was found with other variables can be added to the chart, but the arrows indicating causal relationships will not connect these variables.

[0083] Observing specific examples of the results derived from causal relationships, the following causal relationships can be visually presented: the causal relationship that the concentration of phosphate in the main step affects the total sugar content; the causal relationship that the total sugar content in the main step affects the product concentration in the main step; and the causal relationship that the product concentration in the main step affects the yield of the final product.

[0084] On the other hand, the causal relationship derivation results described above can be input into the generative AI-assisted system 200, and a literature-based explanation of the results can be generated and provided from the generative AI. For example, regarding the causal relationship that phosphate concentration in the main step affects total sugar content, the generative AI can receive evidence such as "phosphate is an essential element for energy metabolism such as ATP synthesis. When the phosphate concentration is sufficient, microorganisms can metabolize sugar more efficiently to increase total sugar content. Furthermore, phosphate can play an important regulatory role in enzyme activation," and a source indicating that this evidence is supported by "studies on the interaction between phosphate metabolism and sugar metabolism" is also provided. That is, the results generated by the causal relationship derivation unit 110 (causal relationship derivation results) can be input into the generative AI-assisted system 200 to generate evidence of the causal relationship and source-related information. This evidence and source-related information can also be provided to the user to help understand or determine variables suitable for process optimization.

[0085] Refer again Figure 3 The causal relationship derivation result calculated by the causal relationship derivation unit 110 can be passed to the causal relationship inference unit 120 (S103), where causal relationship inference calculations can be performed (S104). Specifically, the causal relationship inference unit 120 can perform the function of quantitatively evaluating the strength of the relationship between various variables by accepting the previously generated causal relationship derivation result as input. In addition to the causal relationship derivation result, prior knowledge data and / or AI analysis data generated by generative AI (data based on literature) can also be input into the causal relationship inference unit 120, which can refer to this data to calculate the strength of the causal relationship between variables. When inferring causal relationships, the strength of the causal relationship can be obtained using a statistical inference model, or the conditional probability distribution between variables can be modeled using a probability-based inference model, thereby obtaining the strength of the causal relationship.

[0086] Figure 5 It is a chart that visually displays the results of causal reasoning. Referring to the accompanying drawings, it can be confirmed that the causal reasoning results, previously shown as causal derivation results, also include values ​​representing the strength of the causal relationship between variables. From this result, it can be seen that the causal derivation unit 110 and the causal reasoning unit 120 can also be implemented by a combined structure.

[0087] On the other hand, the causal reasoning result calculated by the causal reasoning unit 120 can be passed to the simulation unit 130 (S105). The simulation unit 130 can be implemented to simulate the process by referring to the causal relationship analyzed (S106) and predict its result.

[0088] The data input into the simulation unit 130 may include simulation condition information in addition to the causal reasoning results (including information related to causal relationships and their strength) previously transmitted from the causal reasoning unit 120. For example, it may also include instructions on which variables to change and how to perform the simulation, such as changing variable 1 from 10 to 20 to perform the simulation, or changing variable 2 from 20 to 30 to perform the simulation. This simulation condition information may be input by a user or automatically generated within the process optimization system 10. In the latter case, the process optimization system 10 can arbitrarily determine the variables whose values ​​need to be changed and the degree of change based on the causal reasoning results, and can be implemented to generate simulation condition information from these determined variables and the degree of change. Even within the process optimization system 10, the generative AI-assisted system 200 can act as the main body for generating the simulation condition information. That is, the generative AI can determine the variables and the degree of change suitable for verifying the causal reasoning results, thereby automatically generating the simulation condition information.

[0089] The data output by the simulation unit 130 includes the expected results when the variable values ​​are changed based on the previously input data. For example, it may include results such as increasing variable 3 (yield) by 5% when variable 1 is changed from 10 to 20, or decreasing variable 3 (yield) by 3% when variable 2 is changed from 20 to 30.

[0090] On the other hand, the results calculated by the simulation unit 130, i.e., the simulation results, can be input into the generative AI-assisted system 200 for the generative AI to generate literature-based data. The generative AI can generate the following information in text form and provide it to the user, such as which literature the simulation results are derived from, whether the simulation results are reasonable when based on various literatures, and whether there are variables that need to be further considered when performing the simulation.

[0091] On the other hand, the simulation results calculated by the simulation unit 130 can be passed to the hypothesis generation unit 140 (S107), so that the hypothesis generation unit 140 can perform hypothesis generation calculations (S108).

[0092] The hypothesis generation unit 140 can determine causal relationships with the potential to improve the process and their corresponding effects by referring to information obtained from simulation results, and can generate a hypothesis that can achieve process optimization from such determined causal relationships. Regarding whether a specific causal relationship has the potential to improve the process, a predefined formula can be used, in which at least one of the following values ​​can be used as a factor: a value for evaluating the effectiveness of the causal relationship; a value for evaluating the probability of being guided to improve the process when the causal relationship occurs effectively; and a value for evaluating the process improvement rate. On the other hand, regarding the hypothesis generated by the hypothesis generation unit 140, it is preferable to generate it based on at least one of the relationships between previously unknown variables or variables that have not been confirmed. As needed, the hypothesis generation unit 140 can request the generative AI-assisted system 200 to verify multiple candidate hypotheses. In this process, the generative AI can determine whether each hypothesis is previously known and whether it contains a previously known causal relationship by querying a literature database, and generate and provide the results in text form. That is, the hypothesis generation unit 140 can receive AI analysis data about at least one hypothesis from the generative AI-assisted system 200.

[0093] The above is for reference only. Figure 3 The process optimization method provided in the first embodiment of the present invention is introduced.

[0094] Figure 6 The process optimization method provided by the second embodiment of the present invention is shown in sequence.

[0095] Referring to the accompanying drawings, in the second embodiment, the causal reasoning unit 120 transmits the causal reasoning result to the hypothesis generation unit 140 instead of transmitting it to the simulation unit 130 (S205). This is different from the first embodiment, although the preceding steps are substantially the same as those in the first embodiment.

[0096] In the second embodiment, after the hypothesis generation unit 140 receives the causal reasoning result, it can first perform a hypothesis generation operation (S206) and then proceed to the step S207 of selecting the hypothesis required for simulation. Step S207 can be accomplished by the user personally researching the generated hypothesis and then selecting a specific hypothesis, or it can be implemented by having a generative AI reference database verify or evaluate the generated hypothesis to select a specific hypothesis.

[0097] On the other hand, after selecting a hypothesis, information about the selected hypothesis can be transmitted to the simulation unit 130 (S208), and simulation calculations can be performed (S209).

[0098] Although the subsequent steps are not illustrated separately, the simulation results generated by the simulation calculation can be processed into a final hypothesis for optimizing the process and provided to the user, or the validity of the above hypothesis can be evaluated again by further verifying the above simulation results (preferably verification based on generative AI).

[0099] The above describes the process optimization method and system for the method provided by the present invention. However, the present invention is not limited to the specific embodiments and application examples described above. Obviously, those skilled in the art can make various modifications without departing from the spirit of the invention as defined in the claims, and such modifications should not be understood separately from the technical concept or prospects of the present invention.

Claims

1. A process optimization method for performing process optimization by a process optimization system including a central processing unit and a memory, the method comprising: (a) a process data receiving step of receiving a plurality of process data; (b) a causality deriving step of analyzing the process data to derive a causality between variables; (c) a causality reasoning step of calculating a strength of the causality between the variables with reference to the derived causality; (d) a simulation step of performing simulation of changing a condition of a specific variable with reference to a causality reasoning result calculated in the causality reasoning step; and (e) a hypothesis generating step of generating at least one or more hypotheses for process optimization with reference to a simulation result calculated in the simulation step.

2. The process optimization method according to claim 1, wherein the causality deriving step includes the steps of: retrieving the process data in order to grasp a feature or a pattern of the process data; deriving a correlation between arbitrary variables included in the process data; selecting a primary variable among the arbitrary variables; and deriving a causality between variables based on the primary variable.

3. The process optimization method according to claim 1, wherein the causality deriving step includes the steps of: inputting a causality derivation result generated by performing the causality deriving step to a generative AI assistance system; and receiving AI analysis data corresponding to the causality derivation result, the AI analysis data being generated by the generative AI assistance system.

4. The process optimization method according to claim 3, wherein the causality deriving step further includes a step of re-deriving the causality between the variables by analyzing the received process data and the AI analysis data.

5. The process optimization method according to claim 4, wherein the AI analysis data includes a causality explanation generated with reference to a stored literature content and a reference related to the literature.

6. The process optimization method according to claim 1, wherein in the causality deriving step, at least one of experimental data or prior knowledge data is input in addition to the process data to derive the causality.

7. The process optimization method according to claim 2, wherein in the step of retrieving the process data, a final index targeted for process optimization is associated with arbitrary variables in the process data to perform the retrieval.

8. The process optimization method according to claim 1, wherein the causality reasoning step includes the steps of: inputting a causality reasoning result generated by performing the causality reasoning step to a generative AI assistance system; and receiving AI analysis data corresponding to the causality reasoning result, the AI analysis data being generated by the generative AI assistance system.

9. The process optimization method according to claim 8, wherein ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ The cause-effect relationship inference step further includes a step of recalculating the strength of the cause-effect relationships between the received cause-effect relationship derivation results and the AI analysis data.

10. The process optimization method according to claim 1, wherein, in the simulation step, in the simulation step, with reference to the cause-effect relationships and the strengths of the cause-effect relationships between specific variables included in the cause-effect relationship inference results, and with reference to simulation condition information arbitrarily generated by the process optimization system or input by any user, simulation is performed.

11. The process optimization method according to claim 1, wherein, in the hypothesis generation step, in the hypothesis generation step, with reference to the simulation results, a hypothesis including a specific cause-effect relationship and an effect caused by the cause-effect relationship is generated, and it is determined whether the hypothesis contains a known cause-effect relationship by searching a literature database.

12. The process optimization method according to claim 11, wherein, the hypothesis generation step includes the following steps: at least one or more hypotheses generated by performing the hypothesis generation step are input to a generative AI assistance system; and AI analysis data corresponding to the hypotheses is received, the AI analysis data being generated by the generative AI assistance system.

13. A process optimization method for performing process optimization by a process optimization system including a central processing unit and a memory, the method comprising: (a) a process data reception step of receiving a plurality of process data; (b) a cause-effect relationship derivation step of deriving cause-effect relationships between variables by analyzing the process data; (c) a cause-effect relationship inference step of calculating strengths of the cause-effect relationships between the variables with reference to the derived cause-effect relationships; (d) a hypothesis generation step of generating one or more hypotheses for process optimization with reference to cause-effect relationship inference results calculated in the cause-effect relationship inference step; and (e) a simulation step of performing simulation of changing conditions of specific variables with reference to information on the hypotheses.

14. The process optimization method according to claim 13, wherein, in a case where a plurality of hypotheses are generated in the hypothesis generation step, a step of selecting a hypothesis required for simulation is further included after the hypothesis generation step, and in the simulation step, only the selected hypothesis is simulated.

15. A process optimization system including a central processing unit and a memory, the central processing unit executing instructions stored in the memory for performing a process optimization method, the process optimization method comprising: (a) a process data reception step of receiving a plurality of process data; (b) a cause-effect relationship derivation step of deriving cause-effect relationships between variables by analyzing the process data; (c) a cause-effect relationship inference step of calculating strengths of the cause-effect relationships between the variables with reference to the derived cause-effect relationships; (d) a simulation step of performing simulation of changing conditions of specific variables with reference to cause-effect relationship inference results calculated in the cause-effect relationship inference step; and ​ ​ (e) a hypothesis generation step of generating at least one or more hypotheses for process optimization with reference to the simulation result calculated in the simulation step.

16. A process optimization system including a central processing unit and a memory, the central processing unit executing instructions stored in the memory for executing a process optimization method, characterized by, the method of executing the process optimization method includes: (a) a process data reception step of receiving a plurality of process data; (b) a causal relationship derivation step of analyzing the process data and deriving a causal relationship between variables; (c) a causal relationship reasoning step of calculating a strength of the causal relationship between the variables with reference to the derived causal relationship; (d) a hypothesis generation step of generating at least one or more hypotheses for process optimization with reference to the causal relationship reasoning result calculated in the causal relationship reasoning step; and (e) a simulation step of performing a simulation of changing a condition of a specific variable with reference to information on the hypothesis.

Images