Formula process optimization method, device and equipment based on multiple agents

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a multi-agent-based formula process optimization method, device and equipment, and the method comprises the steps: collecting production environment parameters in real time through a perception layer agent cluster, generating an environment sensing data set, inputting the environment sensing data set to a digital twinborn, mapping the environment sensing data set into virtual environment state data, and carrying out the simulation of the virtual environment state data. And processing the virtual environment state data through a decision-making layer intelligent agent cluster to obtain process adjustment parameters and a formula topological structure, thereby performing game optimization, and obtaining a target process instruction and a target formula. The environment sensing data set is obtained through the sensing layer intelligent agent to ensure the timeliness and accuracy of the data, virtual simulation of the production process is achieved by mapping the virtual environment state data, process adjustment parameters and a formula topological structure are optimized through game optimization, automation and intelligentization from data collection to optimization decision making are achieved, and the production efficiency is improved. Manual intervention is reduced, and the optimization efficiency of the formula process is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a recipe process optimization method, device and equipment based on multiple agents. BACKGROUND

[0002] In industrial production, recipe process optimization is a key link to improve product quality and efficiency. With the increasing complexity of production processes, higher requirements are placed on the accuracy and adaptability of recipe processes. However, existing optimization methods mostly rely on experience and traditional algorithms, and when faced with massive production data and complex environmental changes, traditional methods are difficult to meet the pursuit of rapid iteration and high performance, resulting in prolonged research and development cycle. SUMMARY

[0003] The main purpose of the present application is to provide a recipe process optimization method, device and equipment based on multiple agents, aiming to solve the technical problem of low efficiency of existing recipe process optimization methods.

[0004] To achieve the above-mentioned purpose, the present application provides a recipe process optimization method based on multiple agents, which comprises:

[0005] The perception layer agent cluster collects production environment parameters in real time to generate environment sensing data sets, wherein the perception layer agent cluster is a distributed perception network constructed by multiple sensors, data acquisition software and signal processing algorithms;

[0006] The environment sensing data set is input into a digital twin to be mapped into virtual environment state data;

[0007] The decision layer agent cluster processes the virtual environment state data to obtain process adjustment parameters and recipe topology structure;

[0008] The process adjustment parameters and the recipe topology structure are optimized through game playing to obtain target process instructions and target recipe.

[0009] In an embodiment, the decision layer agent cluster includes process parameter adjustment agents and recipe recombination agents, the virtual environment state data includes environmental change amount and material response characteristics, and the step of processing the virtual environment state data through the decision layer agent cluster to obtain process adjustment parameters and recipe topology structure comprises:

[0010] According to the environmental change amount, the process adjustment parameters are determined by the process parameter adjustment agents;

[0011] Based on the material response characteristics, the recipe topology structure is generated by the recipe recombination agents.

[0012] In an embodiment, the step of generating a formula topology structure by the formula reorganization agent based on the material response characteristics comprises:

[0013] Resolving the temperature sensitivity coefficient and the pressure tolerance threshold in the material response characteristics;

[0014] When the real-time temperature change amount exceeds the temperature sensitivity coefficient, a first topology substructure is output by activating a flame-retardant molecular network generation module;

[0015] And / or, when the real-time pressure change amount exceeds the pressure tolerance threshold, a second topology substructure is output by activating a three-dimensional cross-linking structure construction module;

[0016] The formula topology structure is obtained according to the first topology substructure and / or the second topology substructure;

[0017] When the real-time temperature change amount does not exceed the temperature sensitivity coefficient and the real-time pressure change amount does not exceed the pressure tolerance threshold, a default topology structure is output as the formula topology structure by calling a basic formula template library.

[0018] In an embodiment, the step of obtaining the formula topology structure according to the first topology substructure and the second topology substructure comprises:

[0019] Extracting molecular bond point data in the first topology substructure;

[0020] Determining the spatial topology coordinates of the second topology substructure;

[0021] Verifying the molecular bond point data and the spatial topology coordinates by a molecular dynamics simulation algorithm;

[0022] When the verification is passed, the first topology substructure and the second topology substructure are fused to generate a formula topology structure.

[0023] In an embodiment, the step of obtaining a target process instruction and a target formula by game optimization of the process adjustment parameter and the formula topology structure comprises:

[0024] Mapping the process adjustment parameter and the formula topology structure into a process energy consumption vector and a stability vector;

[0025] Establishing a game payoff matrix according to the process energy consumption vector and the stability vector;

[0026] Solving the optimal solution of the game payoff matrix by a Nash equilibrium algorithm;

[0027] When the formula stability value corresponding to the optimal solution is lower than a preset threshold, a parameter correction suggestion set is obtained by searching a historical optimization path in a knowledge graph.

[0028] According to the parameter correction suggestion set, the game benefit matrix is corrected, and according to the optimal solution obtained by re-solving, target process instructions and target recipes are determined.

[0029] In an embodiment, the step of retrieving a historical optimization path in the knowledge graph to obtain a parameter correction suggestion set comprises:

[0030] The recipe topological structure is split into a material function segment;

[0031] A historical optimization path with the highest similarity to the material function segment is retrieved in the knowledge graph stored in the blockchain;

[0032] The process adjustment parameters in the historical optimization path are aggregated through a federated learning algorithm to generate a parameter correction suggestion set.

[0033] In an embodiment, after the step of processing the virtual environment state data through the decision layer agent cluster to obtain process adjustment parameters and recipe topological structures, the method further comprises:

[0034] The rate of change of the ambient temperature and the frequency of pressure fluctuation are monitored in real time;

[0035] When the rate of change of the ambient temperature exceeds a critical value, the process adjustment parameters are encrypted through a homomorphic encryption algorithm, and the recipe topological structure is segmented through a key fragmentation algorithm;

[0036] And / or, when the frequency of pressure fluctuation exceeds a threshold value, the process adjustment parameters are subjected to noise interference processing, and the recipe topological structure is subjected to quantum resistance encryption processing;

[0037] The processed process adjustment parameters and the processed recipe topological structure are obtained;

[0038] Correspondingly, the step of game-optimizing the process adjustment parameters and the recipe topological structure to obtain target process instructions and target recipes comprises:

[0039] The processed process adjustment parameters and the processed recipe topological structure are game-optimized to obtain target process instructions and target recipes.

[0040] In addition, to achieve the above-mentioned purpose, the present application also proposes a recipe process optimization device based on multiple agents, which comprises:

[0041] The data acquisition module is configured to acquire production environment parameters in real time through a perception layer agent cluster to generate an environment sensing data set.

[0042] The entity mapping module is configured to input the environment sensing data set into a digital twin to map the environment sensing data set into virtual environment state data.

[0043] The data conversion module is configured to process the virtual environment state data through a decision layer agent cluster to obtain process adjustment parameters and a recipe topology structure.

[0044] The fusion optimization module is configured to perform game optimization on the process adjustment parameters and the recipe topology structure to obtain target process instructions and a target recipe.

[0045] In addition, to achieve the above-mentioned purposes, the present application further provides a recipe process optimization device based on multiple agents, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the recipe process optimization method based on multiple agents as described above.

[0046] In addition, to achieve the above-mentioned purposes, the present application further provides a storage medium, which is a computer readable storage medium, and the storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the recipe process optimization method based on multiple agents as described above.

[0047] The technical scheme provided by the present application acquires production environment parameters in real time through a perception layer agent cluster to generate an environment sensing data set, inputs the environment sensing data set into a digital twin to map the environment sensing data set into virtual environment state data, processes the virtual environment state data through a decision layer agent cluster to obtain process adjustment parameters and a recipe topology structure, and performs game optimization on the process adjustment parameters and the recipe topology structure to obtain target process instructions and a target recipe. The present application acquires environment sensing data set through a perception layer agent to ensure the timeliness and accuracy of the data, inputs the data into a digital twin to map the data into virtual environment state data, realizes virtual simulation of the production process, processes the virtual environment state data through a decision layer agent cluster, and then optimizes the process adjustment parameters and the recipe topology structure through a game optimization method to obtain target process instructions and a target recipe, thereby realizing automation and intelligentization from data acquisition to optimization decision, reducing manual intervention, and significantly improving the efficiency of recipe process optimization. BRIEF DESCRIPTION OF DRAWINGS

[0048] The accompanying drawings, which are incorporated into and form a part of the specification, illustrate preferred embodiments of the present application and, together with the description, serve to explain the principles of the application.

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings required by the embodiments or prior art description will be briefly introduced. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative labor.

[0050] Figure 1 A flowchart provided by the embodiment one of the recipe process optimization method based on multi-agent of the present application;

[0051] Figure 2 A flowchart provided by the embodiment two of the recipe process optimization method based on multi-agent of the present application;

[0052] Figure 3 A flowchart provided by the embodiment three of the recipe process optimization method based on multi-agent of the present application;

[0053] Figure 4 A module structure diagram of the recipe process optimization device based on multi-agent of the embodiment of the present application;

[0054] Figure 5 A device structure diagram of the hardware running environment involved in the recipe process optimization method based on multi-agent of the embodiment of the present application.

[0055] The purpose of the present application, functional characteristics and advantages will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0056] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application, and are not used to limit the present application.

[0057] In order to better understand the technical solutions of the present application, the following will be described in detail in combination with the drawings and specific embodiments of the specification.

[0058] The existing recipe process optimization method depends on experience and traditional algorithm. When facing mass production data and complex environmental changes, the traditional method is difficult to meet the pursuit of rapid iteration and high performance, resulting in the extension of the research and development cycle.

[0059] Therefore, in order to overcome the above-mentioned defects, the present application provides a solution, which uses a perception layer agent to obtain an environment sensing data set, ensures the timeliness and accuracy of the data, inputs these data into a digital twin, maps them into virtual environment state data, realizes virtual simulation of the production process, and a decision layer agent cluster processes the virtual environment state data, and then optimizes the process adjustment parameters and the recipe topology structure through a game optimization method to obtain target process instructions and target recipes, thereby realizing automation and intelligentization from data acquisition to optimization decision, reducing manual intervention, and significantly improving the efficiency of recipe process optimization.

[0060] It should be noted that the execution subject of each embodiment of the present application can be a computing service system with data processing, network communication and program running functions, such as an electronic system capable of realizing the above-mentioned functions, a recipe process optimization system based on multiple agents, etc. Hereinafter, the recipe process optimization system based on multiple agents (hereinafter referred to as "system") is taken as an example to describe each embodiment below.

[0061] Based on this, the embodiment of the present application provides a recipe process optimization method based on multiple agents, which refers to Figure 1 , Figure 1 The flowchart of the first embodiment of the recipe process optimization method based on multiple agents of the present application is shown in FIG. 1.

[0062] In this embodiment, the recipe process optimization method based on multiple agents includes steps S10-S40:

[0063] Step S10, real-time collection of production environment parameters by a perception layer agent cluster to generate an environment sensing data set, wherein the perception layer agent cluster is a distributed perception network constructed by multiple sensors, data acquisition software and signal processing algorithms.

[0064] It should be noted that the perception layer agent cluster is a distributed perception network constructed based on multiple types of sensors, data acquisition software and signal processing algorithms. The above-mentioned sensors can include but are not limited to temperature sensors for accurately measuring temperature changes in the production environment, pressure sensors for real-time monitoring of pressure generated during the production process, humidity sensors for detecting environmental humidity levels, gas concentration sensors for measuring the concentration of specific gases in the production environment, and vibration sensors for monitoring the vibration of equipment, etc. These sensors are distributed at key positions in the production environment and can monitor various environmental parameters in real time.

[0065] Further, the data acquisition software is responsible for collecting and preliminarily arranging the raw data collected by the sensors, and the signal processing algorithm is used to filter and amplify the collected data, so as to improve the accuracy and reliability of the data. Among them, the signal processing algorithm used can be an algorithm combining quantum-inspired generative adversarial network and neuromorphic computing. When processing the environmental signals collected by the perception layer agent cluster, the quantum-inspired generative adversarial network can generate and identify signal features, and the neuromorphic computing simulates the structure and function of the biological nervous system, realizes efficient information processing by simulating the behavior of neurons and synapses. After processing, the noise in the raw data is effectively removed, and the key features of the signal are enhanced, and then the processed data is integrated into a structured environmental sensing data set.

[0066] Step S20, inputting the environmental sensing data set into the digital twin and mapping it into virtual environment state data.

[0067] It should be understood that the environmental sensing data set generated by the perception layer agent cluster is input into the digital twin, and according to the preset mapping rules and algorithms, various parameters and states in the actual production environment can be converted into virtual environment state data.

[0068] Among them, the digital twin is a multi-modal dynamic simulation system that constructs a virtual entity interacting with the actual production environment in both directions by fusing real-time sensing data, physical law constraints and cross-scale modeling technology. The digital twin in this embodiment adopts a physical-data dual-driven engine: the physical layer describes the material response law through parameterized partial differential equations; the data layer integrates the LSTM-GAN (Long Short-Term Memory-Generative Adversarial Networks) trained by federated learning to predict process parameter drift, and verifies the conflict between physical and data models through quantum Monte Carlo algorithm to ensure simulation reliability. In addition, a dynamic knowledge graph is introduced to realize cross-domain experience transfer (such as new energy battery formula optimization transfer to chemical industry), and the simulation path is stored based on blockchain, supporting zero-knowledge proof verification to prevent data tampering.

[0069] Step S30, processing the virtual environment state data through the decision layer agent cluster to obtain process adjustment parameters and formula topology structure.

[0070] Further, the decision layer agent cluster processes the virtual environment state data generated by the digital twin. The decision layer agent cluster is composed of multiple agents with certain decision-making and analysis capabilities. According to the virtual environment state data, the decision layer agent cluster analyzes and calculates the process parameters and formula structure in the production process by combining pre-set rules, models and algorithms. For example, the decision layer agent cluster can obtain process adjustment parameters for the current production environment state, such as adjusting the values of process parameters such as temperature, pressure and flow rate, and the formula topology structure, i.e., determining the connection mode and structure form of each component in the formula, to provide specific process adjustment and formula optimization suggestions for the subsequent production process.

[0071] In step S40, the process adjustment parameters and the formula topology structure are optimized through game playing to obtain target process instructions and target formula.

[0072] It should be noted that game playing optimization is an optimization method that finds the optimal solution by simulating the competition and cooperation process between multiple participants.

[0073] In this step, the process adjustment parameters and the formula topology structure are regarded as participants in the game. By setting certain game rules and objective functions, such as maximizing production efficiency and optimizing product quality, they are allowed to optimize the game under certain constraints, i.e., constantly adjusting and optimizing the process adjustment parameters and the formula topology structure, so as to achieve the optimal performance and effect under the premise of meeting the production requirements and quality standards. After game playing optimization, the target process instructions and the target formula are finally obtained, which will be applied to the actual production process to guide production operations and achieve optimization of the production process and improvement of product quality.

[0074] For example, in the production of lithium battery separators, the process parameters tend to increase the sintering temperature to shorten the molding time, while the formula structure needs to control the temperature fluctuation to avoid material phase change. Through Nash equilibrium game playing, the system automatically converges to the optimal solution (e.g., temperature set to 235°C and material porosity adjusted to 40%), and finally outputs process instructions and formulas that can simultaneously consider production speed and product performance.

[0075] In this embodiment, the perception layer agent acquires the environmental sensing data set to ensure the timeliness and accuracy of the data, inputs these data into the digital twin, maps them into virtual environment state data, realizes virtual simulation of the production process, the decision layer agent cluster processes the virtual environment state data, and then optimizes the process adjustment parameters and the formula topology structure through game playing optimization method to obtain target process instructions and target formula, thereby realizing automation and intelligentization from data acquisition to optimization decision-making, reducing manual intervention, and significantly improving the efficiency of formula process optimization.

[0076] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as the above embodiment one can refer to the above introduction, and the subsequent will not be described. On this basis, referring to Figure 2 , the step S30 can include steps S301-S302:

[0077] Step S301, according to the environmental change, the process adjustment parameters are determined by the process adjustment agent.

[0078] It should be noted that the decision layer agent cluster can include process parameter adjustment agent and formula reorganization agent. The process parameter adjustment agent is an autonomous optimization module based on reinforcement learning and physical modeling, which dynamically generates the optimal adjustment strategy by real-time analysis of the multi-dimensional correlation of process parameters such as temperature, pressure, flow, etc.; the formula reorganization agent is a calculation unit that combines material gene coding and topology optimization algorithm, which reconstructs the molecular-level topology structure of formula components (such as crosslinking density, functional group distribution) according to environmental sensing data and performance indicators.

[0079] In addition, the decision layer agent cluster can also include abnormal diagnosis agent, energy efficiency management agent and other collaborative modules, which together constitute a closed-loop optimization system. The abnormal diagnosis agent identifies abnormal fluctuations (such as temperature mutation, impurity precipitation) in the production process by real-time monitoring of the dynamic correlation of process parameters and formula structure, and triggers self-healing strategies (such as emergency speed reduction, impurity chelating molecule release); the energy efficiency management agent dynamically optimizes equipment operating parameters (such as voltage gradient, stirring frequency) based on the game relationship between process energy consumption and material response characteristics, reduces unit energy consumption under the premise of ensuring quality.

[0080] The virtual environment state data covers the information of environmental change and material response characteristics. The environmental change refers to the change of various parameters in the production environment, such as the increase or decrease of temperature, the fluctuation of pressure, the change of humidity, etc.; the material response characteristics involve the performance and change law of various materials used in the production process under different environmental conditions, such as the change of thermal expansion coefficient, elastic modulus, yield strength of materials with temperature, pressure, etc., as well as the interaction and reaction characteristics between materials.

[0081] It should be noted that, in order to realize efficient cooperation and closed-loop optimization of different intelligent agents in the decision layer cluster, the system deploys a heterogeneous intelligent agent dynamic semantic coupling communication framework, which goes beyond the traditional static JSON or pure text exchange and adopts a multi-modal real-time stream protocol: for high-frequency, low-dimensional process parameters (temperature, pressure, flow) and their multi-dimensional correlation analysis results, millisecond-level pushing is performed through compact binary format time series data stream to ensure the timeliness of the parameter adjustment instruction generation; for the complex structured data required for formula reorganization (such as molecular topology description, performance index constraint, environmental sensing aggregation information), a structured data packaging format with metadata description capability is used for packaging, which embeds lightweight domain ontology mapping information outside the core data payload, so that the receiver can accurately analyze the heterogeneous data model (such as physical model parameters vs. material gene encoding). The strategy feedback (optimal adjustment strategy) and structure reconstruction instruction (molecular-level topology parameter) between the two are interacted bidirectionally through a low-delay instruction channel, which supports event-triggered incremental semantic update and only transmits the change amount and associated context, thereby maximizing the reduction of communication overhead.

[0082] The above-mentioned real-time communication paradigm that integrates streaming, semantic mapping and incremental update ensures the synchronization and consistency of the cluster in making cross-physical-chemical domain decisions in dynamic industrial scenarios.

[0083] In this embodiment, the process parameter adjustment intelligent agent first receives the environmental change amount information in the virtual environment state data, analyzes and evaluates the environmental change amount. Through a pre-set algorithm and model, such as a machine learning-based prediction model or a simulation model based on physical principles, the impact of environmental changes on the production process and product quality is predicted. According to this prediction result, the process parameter adjustment intelligent agent calculates the corresponding process adjustment parameters, such as adjusting the temperature set value of the heating furnace, changing the pressure control range of the reaction kettle, adjusting the speed of the conveyor belt, etc., to ensure that the production process can still run stably under the condition of environmental changes, and the product quality meets the requirements.

[0084] Step S302, based on the material response characteristics, a formula topology structure is generated by a formula reorganization intelligent agent.

[0085] The formula reorganization intelligent agent receives the material response characteristic information in the virtual environment state data, uses its deep understanding and analysis capability of material characteristics, and combines the production target and product quality requirement to optimize and reorganize the various components in the formula. By considering the compatibility, reactivity, performance complementarity and other factors between materials, as well as the response characteristics of materials under different environmental conditions, the formula reorganization intelligent agent uses specific algorithms and models, such as topology optimization algorithms or combination optimization methods based on genetic algorithms, to determine the optimal combination mode and proportion relationship of each component in the formula, thereby generating a new formula topology structure.

[0086] As an implementation, the step S402 in the embodiment can include: analyzing a temperature sensitivity coefficient and a pressure tolerance threshold in the material response characteristics; when a real-time temperature change amount exceeds the temperature sensitivity coefficient, outputting a first topological substructure by activating a flame-retardant molecular network generation module; and / or when a real-time pressure change amount exceeds the pressure tolerance threshold, activating a three-dimensional cross-linked structure construction module to output a second topological substructure; obtaining a formula topological structure according to the first topological substructure and / or the second topological substructure; and when the real-time temperature change amount does not exceed the temperature sensitivity coefficient and the real-time pressure change amount does not exceed the pressure tolerance threshold, calling a basic formula template library to output a default topological structure as the formula topological structure.

[0087] In a specific implementation, first, the formula reorganization agent analyzes key parameters in the material response characteristics. The temperature sensitivity coefficient refers to the sensitivity of material performance to temperature changes; the pressure tolerance threshold refers to the maximum pressure value that the material can withstand, and the performance of the material may change irreversibly beyond this value. The formula reorganization agent accurately extracts these key parameters by analyzing the physical and chemical properties of the material, experimental data, and historical production records.

[0088] In the production process, the real-time temperature change amount is monitored. When the real-time temperature change amount exceeds the temperature sensitivity coefficient of the material, the formula reorganization agent triggers the flame-retardant molecular network generation module. Based on the flame-retardant characteristics of the material, the module generates a first topological substructure with flame-retardant function through specific chemical reactions or physical processes. This substructure can effectively suppress the combustion reaction in a high-temperature environment and improve the fireproof performance of the product. The generation of the first topological substructure is realized through precise chemical formula adjustment and molecular structure design, ensuring the stability and safety of the material performance under high-temperature conditions.

[0089] Meanwhile or separately, when the real-time pressure change amount exceeds the pressure tolerance threshold of the material, the formula reorganization agent activates the three-dimensional cross-linked structure construction module. The module constructs a three-dimensional cross-linked structure with high strength and shock resistance through chemical cross-linking reactions or physical methods. This structure can effectively enhance the mechanical properties of the material under high pressure, preventing the material from deforming or being damaged under pressure. The generation of the second topological substructure is based on the molecular structure and cross-linking reaction characteristics of the material, realized through precise chemical formula and process parameter control, ensuring the stability and reliability of the material under high pressure conditions.

[0090] The formula reorganization agent generates a final formula topology structure according to the generated first topology substructure and / or the second topology substructure, taking into account the performance requirements of the material and the production environment conditions. If both the first and second topology substructures exist, they are reasonably combined and optimized to form a comprehensive topology structure to meet the multiple requirements for material performance in complex environmental conditions. If only one of the substructures exists, it is directly used as the formula topology structure.

[0091] In addition, when the real-time temperature change does not exceed the temperature sensitivity coefficient of the material, and the real-time pressure change does not exceed the pressure tolerance threshold of the material, the production environment is in a relatively stable and normal state. In this case, the formula reorganization agent calls the pre-stored basic formula template library and directly outputs the default topology structure as the formula topology structure. The basic formula template library contains formula topology structures that have been verified and have stable performance under normal production conditions, which can ensure the quality and performance of the product under conventional environmental conditions.

[0092] In addition, it should be noted that in the present embodiment, a temperature-sensitive response function γ = α·e (-β / T) and a trigger topology reorganization condition function In the formula, α is the response intensity coefficient of the base material; e is the environmental response factor; (-β / T) is the process intensity coefficient; T is the real-time temperature; and σ is the pressure breakthrough threshold.

[0093] The temperature-sensitive response function quantifies the sensitivity of the material to temperature changes through an exponential relationship. When the real-time temperature T fluctuates beyond the preset threshold, the formula agent starts the fire-retardant network reconstruction. The pressure trigger function is based on the mechanical properties of the material and activates the molecular chain reorganization of the three-dimensional cross-linked structure when the pressure gradient breaks through the threshold ΔP.

[0094] For example, when a temperature rise of 15℃ is detected, the process parameter adjustment agent immediately responds by reducing the working voltage by 0.3V to stabilize the reaction process, while the formula reorganization agent simultaneously starts the self-assembly mechanism to form a fire-retardant network structure with temperature self-adaptive characteristics. When the sensing layer detects the presence of impurity ions in the electrolyte, the formula agent quickly releases specific chelating molecules to construct a selective ion sieve topology cage through molecular recognition, effectively isolating the impurity interference. In the case where the pressure fluctuation exceeds the preset threshold, the system automatically activates the topology reorganization algorithm to promote the reorganization of the material molecular chain into a three-dimensional cross-linked anti-seismic structure with energy dissipation characteristics, thereby ensuring the performance stability of the product under extreme environmental conditions.

[0095] As an implementation, the step of obtaining a formula topology structure according to the first topology substructure and the second topology substructure can include: extracting molecular bonding site data in the first topology substructure; determining spatial topology coordinates of the second topology substructure; verifying the molecular bonding site data and the spatial topology coordinates through a molecular dynamics simulation algorithm; and when the verification passes, fusing the first topology substructure and the second topology substructure to generate a formula topology structure.

[0096] In a specific implementation, after the formula reorganization agent generates the first topology substructure, it needs to extract the molecular bonding site data in the first topology substructure for information analysis of the substructure. Molecular bonding site data refers to the position information of a molecule that can form a chemical bond with other molecules or atoms. It can include the type of bonding site (such as covalent bond, ionic bond, etc.), position coordinates, bond angle, and bond energy, etc.

[0097] For the second topology substructure, the formula reorganization agent needs to determine its spatial topology coordinates. Spatial topology coordinates refer to the specific positions of atoms or molecular groups in a molecule in three-dimensional space. These coordinate information is obtained through precise molecular modeling and calculation, which can reflect the arrangement and distribution of molecules in space. The formula reorganization agent uses molecular simulation software and mathematical algorithms to calculate the spatial coordinates of each key point in the second topology substructure, ensuring that in the subsequent fusion process, the two substructures can be accurately matched and connected in space.

[0098] Before fusing the two topology substructures, the extracted molecular bonding site data and the determined spatial topology coordinates need to be verified. Molecular dynamics simulation algorithm is a calculation method based on physical principles, which can simulate the motion and interaction of molecules under different conditions. The formula reorganization agent uses this algorithm to simulate and verify the molecular bonding site data of the first topology substructure and the spatial topology coordinates of the second topology substructure. By simulating the bonding process and spatial arrangement of molecules, it checks whether there are conflicts in chemical bonding, steric hindrance, etc. Only when the simulation results show that the two substructures are feasible in chemical bonding and spatial arrangement, the verification is passed.

[0099] When the molecular dynamics simulation algorithm is verified, the formula reorganization agent fuses the first topology substructure and the second topology substructure, that is, through precise chemical bonding and spatial arrangement adjustment, the two substructures are organically combined together to form a complete formula topology structure.

[0100] In addition, during verification, comprehensive verification can also be performed through various calculation methods. Specifically, after extracting the molecular bonding site data of the first topological substructure and determining the spatial topological coordinates of the second topological substructure, a molecular dynamics model can be used to provide a more bottom-level theoretical framework and parameterization basis, a finite element simulation can be used to analyze the structural stability, stress distribution and other macroscopic physical properties, a high-throughput calculation method can be used to quickly screen the optimal binding site and conformation, and a molecular dynamics simulation algorithm can be used to simulate the dynamic binding process. Various complementary methods cross-verify these data to ensure the compatibility and binding feasibility of the bonding site and spatial coordinates. Only after verification, the formula reorganization agent will perform fusion operation to integrate the two substructures organically through accurate regulation of chemical bonding and spatial arrangement, and finally generate a complete and multi-dimensionally verified formula topological structure.

[0101] The embodiment can better adapt to the dynamic changes of the production environment and further improve the flexibility and adaptability of formula process optimization by determining the process adjustment parameters through the process parameter adjustment agent according to the environmental change amount and generating the formula topological structure through the formula reorganization agent based on the material response characteristics.

[0102] Based on the first embodiment of the present application, the same or similar contents as the above embodiment one can be referred to the above introduction, and will not be described in detail hereinafter. On this basis, please refer to Figure 3 , the step S40 can include steps S401-S405:

[0103] Step S401, mapping the process adjustment parameters and the formula topological structure into process energy consumption vectors and stability vectors.

[0104] It should be understood that the embodiment first needs to convert the process adjustment parameters and the formula topological structure into a quantifiable form. The process adjustment parameters mainly involve the energy consumption in the production process, which are mapped into process energy consumption vectors, while the formula topological structure is closely related to the stability of the product, which is mapped into stability vectors. The mapping method can adopt a dynamic weight distribution model based on attention mechanism to adapt to the parameter sensitivity in different production scenarios.

[0105] Step S402, establishing a game payoff matrix according to the process energy consumption vectors and the stability vectors.

[0106] Further, after obtaining the process energy consumption vectors and the stability vectors, a game payoff matrix needs to be constructed next. The game payoff matrix is a mathematical tool for quantifying the multi-objective optimization relationship between the process energy consumption vectors and the stability vectors into a two-dimensional matrix. The rows and columns of the matrix respectively represent different game strategies of process parameter combinations and formula structures, and the matrix element values are calculated by Nash equilibrium algorithm. The comprehensive revenue score (such as energy consumption reduction rate x stability gain coefficient) under each strategy combination.

[0107] At step S403, the optimal solution of the game payoff matrix is solved by Nash equilibrium algorithm.

[0108] It should be noted that Nash equilibrium is a concept in game theory, which means that each participant has chosen the optimal strategy, provided that the strategies of other participants remain unchanged. In this step, Nash equilibrium algorithm is used to solve the game payoff matrix to find the optimal solution, which represents the most ideal process parameters and formulation structure combination under the current trade-off between energy consumption and stability.

[0109] At step S404, when the formulation stability value corresponding to the optimal solution is lower than the preset threshold, the historical optimization path is retrieved in the knowledge graph to obtain a parameter correction suggestion set.

[0110] It should be understood that if the optimal solution obtained by Nash equilibrium algorithm corresponds to a formulation stability value lower than the preset threshold, it means that the current decision scheme has deficiencies in stability. At this time, the knowledge graph needs to be used to find a solution. Knowledge graph is a structured knowledge representation method, which stores rich historical optimization paths and experience data. By retrieving similar historical cases in the knowledge graph, a series of parameter correction suggestions can be obtained. The suggestion set contains experience and methods on how to adjust process parameters and formulation structure to improve stability in similar situations.

[0111] As an implementation, the above step S404 in the embodiment can include: splitting the formulation topology structure into material function fragments; retrieving the historical optimization path with the highest similarity to the material function fragments in the knowledge graph stored in the blockchain; and aggregating the process adjustment parameters in the historical optimization path by federated learning algorithm to generate a parameter correction suggestion set.

[0112] In specific implementation, the formulation topology structure needs to be analyzed and split in detail first. The formulation topology structure contains multiple materials and their interactions. Splitting it into material function fragments means that the entire structure needs to be decomposed into several function fragments according to the physical and chemical properties, functional characteristics and role in the formulation. For example, a formulation can contain flame-retardant function fragments, reinforcing function fragments, and stability function fragments.

[0113] After the recipe topology is split into material function segments, the next step is to search in the knowledge graph stored in the blockchain. The blockchain technology provides a secure and tamper-proof storage environment for the knowledge graph, ensuring the authenticity and reliability of the data in the knowledge graph. Through specific search algorithms, such as semantic similarity-based search algorithms, the knowledge graph is queried to find the historical optimization path with the highest similarity to the current material function segment, and then the process adjustment parameters in these paths are aggregated.

[0114] It should be noted that the federated learning algorithm is a distributed machine learning method that can jointly learn data scattered in different nodes while protecting data privacy. In this step, the federated learning algorithm is used to aggregate the process adjustment parameters in the retrieved historical optimization paths, and the parameters in multiple historical optimization paths are analyzed comprehensively to generate a comprehensive parameter correction suggestion set.

[0115] Step S405, according to the parameter correction suggestion set, the game payoff matrix is corrected, and according to the optimal solution obtained by re-solving, the target process instruction and the target recipe are determined.

[0116] According to the parameter correction suggestion set obtained from the knowledge graph, the game payoff matrix is corrected, that is, some elements in the matrix are adjusted to reflect the new parameter settings and the expected performance improvement. The corrected game payoff matrix is more in line with the actual production demand and stability requirements. Then, the Nash equilibrium algorithm is used again to solve the corrected matrix to obtain a new optimal solution. If the optimization target and / or the number of iterations are reached, this new optimal solution will be the final target process instruction and target recipe.

[0117] This embodiment not only considers the two key factors of process energy consumption and recipe stability, but also further improves the accuracy and stability of the optimization result by introducing the knowledge graph and historical optimization path. The dynamic correction mechanism can better adapt to changes in the production environment.

[0118] As an implementation manner, after the step S30, the method can further include: monitoring the environmental temperature change rate and the pressure fluctuation frequency in real time; when the environmental temperature change rate exceeds a critical value, performing encryption processing on the process adjustment parameters by using a homomorphic encryption algorithm, and performing segmentation processing on the recipe topology by using a key fragmentation algorithm; and / or when the pressure fluctuation frequency exceeds a threshold value, performing noise interference processing on the process adjustment parameters, and performing quantum resistance encryption processing on the recipe topology; obtaining the processed process adjustment parameters and the processed recipe topology.

[0119] Correspondingly, the step of performing game optimization on the processed process adjustment parameters and the processed recipe topology to obtain target process instructions and a target recipe includes: performing game optimization on the processed process adjustment parameters and the processed recipe topology to obtain target process instructions and a target recipe.

[0120] In a specific implementation, after the decision layer agent cluster processes the virtual environment state data to obtain the process adjustment parameters and the recipe topology, the system enters a real-time monitoring phase. The core task of this phase is to continuously track two key parameters in the production environment: the environmental temperature change rate and the pressure fluctuation frequency. The temperature change rate refers to the amount of change in the environmental temperature per unit time, reflecting the dynamic situation of heat transfer and temperature control in the production process; the pressure fluctuation frequency refers to the number of pressure changes per unit time, revealing the stability of the pressure in the production process. By deploying high-precision sensors and data acquisition systems, accurate values of these parameters are obtained in real time.

[0121] When the real-time monitored environmental temperature change rate exceeds the preset critical value, the system starts a security processing mechanism. First, the process adjustment parameters are encrypted using a homomorphic encryption algorithm. Homomorphic encryption allows direct calculation on encrypted data without decrypting the data first, ensuring that the process adjustment parameters remain usable in subsequent processing while protecting data privacy. Second, the recipe topology is segmented using a key fragmentation algorithm. The key fragmentation algorithm divides the key into multiple fragments, which are stored in different locations. Only by combining these fragments can the original key be restored, further enhancing the security of the recipe topology.

[0122] At the same time or separately, when the pressure fluctuation frequency exceeds the preset threshold, the system takes another set of security processing measures. The process adjustment parameters are subjected to noise interference processing by adding appropriate random noise to the data, masking the true data characteristics and reducing the readability of the data during transmission and processing, preventing malicious analysis and use of the data. The recipe topology is subjected to quantum-resistant encryption processing to resist quantum computing attacks and ensure the security of the recipe topology in future quantum computing environments.

[0123] After the above security processing steps, the system obtains the processed process adjustment parameters and the recipe topology. These data, which have been encrypted, segmented, subjected to noise interference, and subjected to quantum-resistant encryption, retain the useful information of the original data while enhancing the security and privacy protection of the data. After obtaining the processed process adjustment parameters and the recipe topology, the processed data is analyzed and optimized using game optimization methods such as Nash equilibrium algorithm to obtain the optimal process adjustment scheme and recipe structure.

[0124] It should be noted that the above examples are only used for understanding the present application and do not constitute a limitation on the multi-agent based formula process optimization method of the present application. More forms of simple changes based on this technical concept are within the protection scope of the present application.

[0125] The present application also provides a multi-agent based formula process optimization device, please refer to Figure 4 , the multi-agent based formula process optimization device comprises:

[0126] The data acquisition module 10 is used for collecting production environment parameters in real time through the perception layer agent cluster to generate environment sensing data sets, wherein the perception layer agent cluster is a distributed perception network constructed by a plurality of sensors, data acquisition software and signal processing algorithms.

[0127] The entity mapping module 20 is used for inputting the environment sensing data set into the digital twin and mapping it into virtual environment state data.

[0128] The data conversion module 30 is used for processing the virtual environment state data through the decision layer agent cluster to obtain process adjustment parameters and formula topology structure.

[0129] The fusion optimization module 40 is used for game optimization of the process adjustment parameters and the formula topology structure to obtain target process instructions and target formula.

[0130] The multi-agent based formula process optimization device provided by the present application adopts the multi-agent based formula process optimization method in the above embodiment, which can solve the technical problem of low efficiency of the existing formula process optimization method. Compared with the prior art, the multi-agent based formula process optimization device provided by the present application has the same beneficial effects as the multi-agent based formula process optimization method provided by the above embodiment, and other technical features in the multi-agent based formula process optimization device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0131] The present application provides a multi-agent based formula process optimization device, which comprises at least one processor and a memory in communication connection with the at least one processor. The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the multi-agent based formula process optimization method in the above embodiment one.

[0132] The following refers to Figure 5This document illustrates a structural schematic diagram of a multi-agent-based formulation and process optimization device suitable for implementing embodiments of this application. The multi-agent-based formulation and process optimization device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Desction), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 5 The multi-agent-based formulation and process optimization device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0133] like Figure 5 As shown, the multi-agent-based formulation process optimization device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the multi-agent-based formulation process optimization device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. The communication device 1009 allows the multi-agent-based formulation and process optimization equipment to communicate wirelessly or wiredly with other devices to exchange data. While the figure shows a multi-agent-based formulation and process optimization equipment with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0134] In particular, according to embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising program code for performing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a read-only memory 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the methods of the embodiments of the present disclosure are performed.

[0135] The multi-agent-based recipe process optimization device provided by the present application adopts the multi-agent-based recipe process optimization method in the above embodiments, and can solve the technical problem of low efficiency of the existing recipe process optimization method. Compared with the prior art, the multi-agent-based recipe process optimization device provided by the present application has the same beneficial effects as the multi-agent-based recipe process optimization method provided by the above embodiments, and other technical features in the multi-agent-based recipe process optimization device are the same as the features disclosed in the previous embodiment method, which will not be repeated here.

[0136] It should be understood that parts of the present disclosure can be realized by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0137] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0138] The present application provides a computer readable storage medium having stored thereon computer readable program instructions (i.e. computer program) for performing the multi-agent-based recipe process optimization method in the above embodiments.

[0139] The computer readable storage medium provided in the application may be, for example, a U disk, but is not limited to an electric, magnetic, optical, electromagnetic, infrared, or semiconductor system or device, or any combination of the above. More specific examples of the computer readable storage medium may include, but are not limited to, an electric connection with one or more conductive wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the embodiment, the computer readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system or device. The program code contained on the computer readable storage medium can be transmitted by any suitable medium, including but not limited to an electric wire, an optical cable, an RF (Radio Frequency), and the like, or any suitable combination of the above.

[0140] The computer readable storage medium described above may be contained in the multi-agent based formula process optimization device, or may exist independently without being assembled into the multi-agent based formula process optimization device.

[0141] The computer readable storage medium described above carries one or more programs, which, when executed by the multi-agent based formula process optimization device, cause the multi-agent based formula process optimization device to: collect production environment parameters in real time through a perception layer agent cluster to generate an environment sensing data set, wherein the perception layer agent cluster is a distributed perception network constructed by a plurality of sensors, data acquisition software, and signal processing algorithms; input the environment sensing data set to a digital twin, map it to virtual environment state data; process the virtual environment state data through a decision layer agent cluster to obtain process adjustment parameters and formula topology structure; and game optimize the process adjustment parameters and formula topology structure to obtain target process instructions and target formula.

[0142] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0143] The computer program instructions can also be loaded onto a computer or other programmable information processing apparatus to cause a series of operations to be performed on the computer or other programmable information processing apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable information processing apparatus implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0144] The modules involved in the embodiments of the present application can be implemented in software or hardware. In some cases, the names of the modules do not constitute a limitation on the modules themselves.

[0145] The readable storage medium provided by the application is a computer readable storage medium, which stores computer readable program instructions (i.e. computer programs) for executing the above-mentioned multi-agent based recipe process optimization method, and can solve the technical problem of low efficiency of the existing recipe process optimization method. Compared with the prior art, the beneficial effects of the computer readable storage medium provided by the application are the same as those of the multi-agent based recipe process optimization method provided by the above-mentioned embodiments, and are not described here.

[0146] The above is only part of the embodiments of the application, and does not limit the protection scope of the application. Any equivalent structural transformation made by using the content of the specification and drawings, or direct / indirect application in other related technical fields under the technical concept of the application is included in the protection scope of the application.

Claims

1. A multi-agent based recipe process optimization method, characterized in that, The method comprises the following steps: Real-time collection of production environment parameters by a perception layer agent cluster to generate an environment sensing data set, wherein the perception layer agent cluster is a distributed sensing network constructed by multiple sensors, data acquisition software and signal processing algorithms; Inputting the environment sensing data set into a digital twin to map it into virtual environment state data; Processing the virtual environment state data by a decision layer agent cluster to obtain process adjustment parameters and a recipe topology structure; Game optimization of the process adjustment parameters and the recipe topology structure to obtain target process instructions and a target recipe.

2. The multi-agent based recipe process optimization method of claim 1, wherein, The decision layer agent cluster comprises a process parameter adjustment agent and a recipe recombination agent, the virtual environment state data comprises an environment change amount and material response characteristics, and the step of processing the virtual environment state data by the decision layer agent cluster to obtain process adjustment parameters and a recipe topology structure comprises: Determining process adjustment parameters by the process parameter adjustment agent according to the environment change amount; Generating a recipe topology structure by the recipe recombination agent based on material response characteristics.

3. The multi-agent based recipe process optimization method of claim 2, wherein, The step of generating a recipe topology structure by the recipe recombination agent based on material response characteristics comprises: Analyzing a temperature sensitivity coefficient and a pressure tolerance threshold in the material response characteristics; When a real-time temperature change amount exceeds the temperature sensitivity coefficient, outputting a first topology substructure by activating a flame-retardant molecular network generation module; And / or, when a real-time pressure change amount exceeds the pressure tolerance threshold, activating a three-dimensional cross-linking structure construction module to output a second topology substructure; Obtaining a recipe topology structure according to the first topology substructure and / or the second topology substructure; When the real-time temperature change amount does not exceed the temperature sensitivity coefficient and the real-time pressure change amount does not exceed the pressure tolerance threshold, calling a basic recipe template library to output a default topology structure as the recipe topology structure.

4. The multi-agent based recipe process optimization method of claim 3, wherein, The step of obtaining a recipe topology structure according to the first topology substructure and the second topology substructure comprises: Extracting molecular bonding point data in the first topology substructure; Determining spatial topology coordinates of the second topology substructure; Verifying the molecular bonding point data and the spatial topology coordinates by a molecular dynamics simulation algorithm; When the verification is passed, fusing the first topology substructure and the second topology substructure to generate a recipe topology structure.

5. The multi-agent based recipe process optimization method of any one of claims 1 to 4, wherein, The step of game optimizing the process adjustment parameters and the recipe topology structure to obtain target process instructions and a target recipe comprises: Mapping the process adjustment parameters and the recipe topology structure into a process energy consumption vector and a stability vector; Establishing a game payoff matrix according to the process energy consumption vector and the stability vector; Solving an optimal solution of the game payoff matrix by a Nash equilibrium algorithm; When a recipe stability value corresponding to the optimal solution is lower than a preset threshold, retrieving a historical optimization path in a knowledge graph to obtain a parameter correction suggestion set; According to the parameter correction suggestion set, correcting the game payoff matrix, and determining target process instructions and a target recipe according to the optimal solution obtained by re-solving.

6. The multi-agent based recipe process optimization method of claim 5, wherein, The step of searching for a historical optimization path in the knowledge graph to obtain a parameter correction suggestion set comprises: The formula topology structure is split into material function segments; In the knowledge graph stored in the blockchain, search for the historical optimization path with the highest similarity to the material function segment; Through the federated learning algorithm, aggregate the process adjustment parameters in the historical optimization path to generate a parameter correction suggestion set.

7. The multi-agent based recipe process optimization method of any one of claims 1 to 4, wherein, After the step of processing the virtual environment state data through the decision layer agent cluster to obtain the process adjustment parameters and the formula topology structure, the method further comprises: Real-time monitoring of environmental temperature change rate and pressure fluctuation frequency; When the environmental temperature change rate exceeds a critical value, the process adjustment parameters are encrypted by a homomorphic encryption algorithm, and the formula topology structure is segmented by a key fragmentation algorithm; And / or, when the pressure fluctuation frequency exceeds a threshold value, the process adjustment parameters are subjected to noise interference processing, and the formula topology structure is subjected to quantum resistance encryption processing; Obtain the processed process adjustment parameters and the processed formula topology structure; Correspondingly, the step of game optimizing the process adjustment parameters and the formula topology structure to obtain target process instructions and target formula comprises: Game optimizing the processed process adjustment parameters and the processed formula topology structure to obtain target process instructions and target formula.

8. A multi-agent based recipe process optimization apparatus, characterized by, The formula process optimization device based on multiple agents comprises: A data acquisition module for acquiring production environment parameters in real time through a perception layer agent cluster to generate an environment sensing data set; An entity mapping module for inputting the environment sensing data set into a digital twin to map it into virtual environment state data; A data conversion module for processing the virtual environment state data through a decision layer agent cluster to obtain process adjustment parameters and formula topology structure; A fusion optimization module for game optimizing the process adjustment parameters and the formula topology structure to obtain target process instructions and target formula.

9. A multi-agent based recipe process optimization apparatus, characterized by, The formula process optimization device based on multiple agents comprises a memory, a processor, and a formula process optimization program based on multiple agents stored on the memory and executable on the processor, wherein the formula process optimization program based on multiple agents, when executed by the processor, implements the formula process optimization method based on multiple agents as claimed in any one of claims 1 to 7.

10. A storage medium, characterized by The storage medium stores a formula process optimization program based on multiple agents, wherein the formula process optimization program based on multiple agents, when executed by the processor, implements the formula process optimization method based on multiple agents as claimed in any one of claims 1 to 7.

Images