Industrial production operation intelligent management method and system based on cooperation of multiple AI agents

Through multi-AI agent collaboration technology, user input is analyzed and tasks are decomposed, and distributed to professional AI agents for processing, which solves the problems of complex data processing and difficult real-time monitoring in industrial production operations, and improves system response efficiency and management efficiency. The accuracy and efficiency of task execution are achieved.

CN119047996APending Publication Date: 2024-11-29SHANGHAI HAO CANG SYST CONTROL TECH CO LTD
View PDF 0 Cites 6 Cited by

Patent Information

Application Number
CN202411006668.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-25
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The existing industrial production operation and management systems have problems such as complex data processing, difficult real-time monitoring, slow response speed and lack of a unified intelligent management platform, resulting in data silos and information asymmetry, which is difficult to meet the needs of modern industrial production.

Method used

Adopt the intelligent management method of industrial production operations based on multi-AI agent collaboration, analyze user input through natural language processing technology, decompose tasks and distribute them to AI agents in professional fields, monitor and execute in real time, and finally integrate them through data aggregation and model inference. Analysis and decision-making ensure smooth information transmission and processing.

Benefits of technology

It effectively solves the problems of complex data processing and difficult real-time monitoring, improves system response efficiency and management efficiency, ensures the accuracy and efficiency of task execution, and optimizes all aspects of production operation management.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119047996A_ABST
    Figure CN119047996A_ABST
Patent Text Reader

Abstract

The invention provides an industrial production operation intelligent management method and system based on multi-AI agent cooperation. The method comprises the steps that 1, a user logs in a terminal equipment application; 2, analyzing a natural language input by a user by using an NLP technology, inferring a specific intention of the user by analyzing a context and an interaction history of the user, and converting the specific intention into a clear task description; step 3, task decomposition and distribution; step 4, executing tasks and feeding back by each sub-agent; and 5, receiving and integrating execution results of the sub-agents through the task coordinator, performing comprehensive analysis and processing on the results, and feeding back a comprehensive final result to a user or a system to complete comprehensive processing and response of task requirements. According to the invention, a layered structure is adopted to effectively solve the problems of complex data processing, large real-time monitoring difficulty and slow system response speed in industrial production operation, so that smooth transmission and processing of information among different levels are ensured, and the overall response efficiency and management efficiency of the system are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an intelligent management method and system for industrial production operations based on collaboration of multiple AI agents. Background Art

[0002] With the rapid development of artificial intelligence (AI) technology, various specialized AI agents have been widely used in various fields. Breakthroughs in AI in visual recognition, speech processing, image analysis, data retrieval, and natural language generation have provided strong technical support for intelligent management and optimization.

[0003] At present, industrial production operation management mainly relies on traditional automation control systems and single intelligent technology. Although these systems have improved production efficiency and management level to a certain extent, they still have problems such as complex data processing, difficulty in real-time monitoring, slow response speed, and lack of a unified intelligent management platform leading to data silos and information asymmetry. With the increasing complexity and diversification of the industrial production environment, the existing management model can no longer meet the needs of modern industrial production. Summary of the invention

[0004] In view of the defects in the prior art, the purpose of the present invention is to provide an intelligent management method and system for industrial production operations based on the collaboration of multiple AI agents.

[0005] The industrial production operation intelligent management method based on multi-AI intelligent agent collaboration provided by the present invention includes:

[0006] Step 1: The user logs in to the terminal device application and enters natural language;

[0007] Step 2: Use natural language processing (NLP) technology to parse the natural language input by the user, including semantic understanding, intent classification, and entity recognition. By analyzing the context and the user's interaction history, the user's specific intent is inferred and the abstract user intent is converted into a clear task description.

[0008] Step 3: Task decomposition and task distribution;

[0009] Step 4: Each sub-agent performs tasks and provides feedback;

[0010] Step 5: Receive and integrate the execution results of each sub-agent through the task coordinator, conduct comprehensive analysis and processing on the results through data aggregation methods, and feed back the comprehensive final results to the user or system to complete the comprehensive processing and response of the task requirements.

[0011] Preferably, step 2 comprises:

[0012] Use natural language processing (NLP) technology to parse the natural language input by the user, specifically:

[0013] Word segmentation: split the text input by the user into a sequence of words or subwords. Word segmentation methods include rule-based word segmentation and statistical-based word segmentation. The maximum matching method is used for Chinese word segmentation.

[0014] Part-of-speech tagging: determining the part of speech of each word in context;

[0015] Entity recognition: Identify entities with specific meanings in text, including names of people, places, and organizations;

[0016] Semantic understanding: converting natural language into structured representations;

[0017] By analyzing the context and the user's interaction history, the user's specific intentions are inferred, specifically:

[0018] Collect the user's current input, context information and interaction history, including: the text currently input by the user, the user's previous interaction records and the user's personal information;

[0019] Use NLP technology to pre-process and parse user input, including: word segmentation, part-of-speech tagging, named entity recognition and syntactic parsing;

[0020] Extract the contextual features of the current conversation, including topics, keywords, and emotions; analyze the user's previous interaction records, including previous queries, click behaviors, and feedback; use context modeling technology to capture the dynamic characteristics of the conversation and long-term and short-term dependencies;

[0021] Train an intent classification model to identify the user's intent, representing the user input, context features, and user history as a feature vector; input the feature vector to the intent classification model, and output the predicted intent category;

[0022] Combine context and historical data, and use implicit intent models to mine users’ implicit intents; process multiple rounds of conversations, and make dynamic adjustments and inferences by tracking changes in conversation status and user intents;

[0023] Transform inferred user intent into clear task descriptions, break down complex user intent into specific subtasks; interact with users to clarify ambiguous intent.

[0024] Preferably, the step 3 comprises:

[0025] Break down large tasks into small tasks based on user needs, and each small task has a clear execution path and goal;

[0026] Determine the execution priority of each subtask based on the task's urgency, resource requirements, and dependencies;

[0027] Each subtask is assigned to the AI ​​sub-agent in the corresponding professional field through the task coordinator.

[0028] Preferably, step 4 comprises:

[0029] Each sub-agent starts to execute the assigned task and monitors the execution progress and status of the task in real time;

[0030] Each sub-agent generates corresponding execution results according to task requirements;

[0031] The sub-agents feed back the execution results to the task coordinator for result integration and comprehensive analysis.

[0032] Preferably, step 5 comprises:

[0033] Data aggregation: Integrate the execution results from different sub-agents for comprehensive analysis and decision-making;

[0034] The data aggregation process is as follows:

[0035] Collect execution results from each sub-agent, including structured data, semi-structured data, and unstructured data;

[0036] {R1,R2,...,R n}

[0037] Among them, R i represents the execution result of the ith sub-agent; n is the number of sub-agents;

[0038] The Excel function Clean() is used to clean and preprocess the collected data to remove redundancy and noise and ensure data quality;

[0039] D clean =Clean(D)

[0040] Where D represents the original data set, D clean represents the cleaned data set;

[0041] The normalization function Normalize() is used to convert data in different formats into a unified format and perform standardization.

[0042] D norm =Normalize(D clean )

[0043] Among them, D norm Represents the standardized data set;

[0044] Use the data merge() function to merge the results of each sub-agent into an aggregate data set D merged , for subsequent analysis and reasoning;

[0045] D merged =Merge(D nirm,1 ,D norm,2 ,...,D norm,n )

[0046] Model reasoning: Use machine learning or deep learning models to reason and make decisions on aggregated data. The model reasoning process is:

[0047] Extract feature vectors from the aggregated dataset for model input;

[0048]

[0049] in, represents the feature extraction function, X represents the feature matrix;

[0050] Choose a machine learning or deep learning model that is appropriate for the task;

[0051] If the model used needs training, use historical data to train the model;

[0052]

[0053] in, represents the loss function, Y represents the label, θ represents the model parameter, θ * are the model parameters after training;

[0054] Use the trained model to make inferences and predictions on new data;

[0055] Y * =f(X;θ * )

[0056] Where f represents the model function, Y * Represents the results predicted by the model.

[0057] The industrial production operation intelligent management system based on multi-AI intelligent agent collaboration provided by the present invention includes:

[0058] Module M1: enables the user to log in to the terminal device application and input natural language;

[0059] Module M2: Use natural language processing (NLP) technology to parse the natural language input by users, including semantic understanding, intent classification, and entity recognition. By analyzing the context and user interaction history, the user's specific intent is inferred and the abstract user intent is converted into a clear task description.

[0060] Module M3: Task decomposition and task distribution;

[0061] Module M4: each sub-agent performs tasks and provides feedback;

[0062] Module M5: Receive and integrate the execution results of each sub-agent through the task coordinator, conduct comprehensive analysis and processing of the results through data aggregation methods, and feed back the comprehensive final results to the user or system to complete the comprehensive processing and response of task requirements.

[0063] Preferably, the module M2 comprises:

[0064] Use natural language processing (NLP) technology to parse the natural language input by the user, specifically:

[0065] Word segmentation: split the text input by the user into a sequence of words or subwords. Word segmentation methods include rule-based word segmentation and statistical-based word segmentation. The maximum matching method is used for Chinese word segmentation.

[0066] Part-of-speech tagging: determining the part of speech of each word in context;

[0067] Entity recognition: Identify entities with specific meanings in text, including names of people, places, and organizations;

[0068] Semantic understanding: converting natural language into structured representations;

[0069] By analyzing the context and the user's interaction history, the user's specific intentions are inferred, specifically:

[0070] Collect the user's current input, context information and interaction history, including: the text currently input by the user, the user's previous interaction records and the user's personal information;

[0071] Use NLP technology to pre-process and parse user input, including: word segmentation, part-of-speech tagging, named entity recognition and syntactic parsing;

[0072] Extract the contextual features of the current conversation, including topics, keywords, and emotions; analyze the user's previous interaction records, including previous queries, click behaviors, and feedback; use context modeling technology to capture the dynamic characteristics of the conversation and long-term and short-term dependencies;

[0073] Train an intent classification model to identify the user's intent, representing the user input, context features, and user history as a feature vector; input the feature vector to the intent classification model, and output the predicted intent category;

[0074] Combine context and historical data, and use implicit intent models to mine users’ implicit intents; process multiple rounds of conversations, and make dynamic adjustments and inferences by tracking changes in conversation status and user intents;

[0075] Transform inferred user intent into clear task descriptions, break down complex user intent into specific subtasks; interact with users to clarify ambiguous intent.

[0076] Preferably, the module M3 comprises:

[0077] Break down large tasks into small tasks based on user needs, and each small task has a clear execution path and goal;

[0078] Determine the execution priority of each subtask based on the task's urgency, resource requirements, and dependencies;

[0079] Each subtask is assigned to the AI ​​sub-agent in the corresponding professional field through the task coordinator.

[0080] Preferably, the module M4 comprises:

[0081] Each sub-agent starts to execute the assigned task and monitors the execution progress and status of the task in real time;

[0082] Each sub-agent generates corresponding execution results according to task requirements;

[0083] The sub-agents feed back the execution results to the task coordinator for result integration and comprehensive analysis.

[0084] Preferably, the module M5 comprises:

[0085] Data aggregation: Integrate the execution results from different sub-agents for comprehensive analysis and decision-making;

[0086] The data aggregation process is as follows:

[0087] Collect execution results from each sub-agent, including structured data, semi-structured data, and unstructured data;

[0088] {R1,R2,...,R n}

[0089] Among them, R i represents the execution result of the ith sub-agent; n is the number of sub-agents;

[0090] The Excel function Clean() is used to clean and preprocess the collected data to remove redundancy and noise and ensure data quality;

[0091] D clean =Clean(D)

[0092] Where D represents the original data set, D clean represents the cleaned data set;

[0093] The normalization function Normalize() is used to convert data in different formats into a unified format and perform standardization.

[0094] D norm =Normalize(D clean )

[0095] Among them, D norm Represents the standardized data set;

[0096] Use the data merge() function to merge the results of each sub-agent into an aggregate data set D merged , for subsequent analysis and reasoning;

[0097] D merged =Merge(D norm,1 ,D norm,2 ,...,D norm,n )

[0098] Model reasoning: Use machine learning or deep learning models to reason and make decisions on aggregated data. The model reasoning process is:

[0099] Extract feature vectors from the aggregated dataset for model input;

[0100]

[0101] in, represents the feature extraction function, X represents the feature matrix;

[0102] Choose a machine learning or deep learning model that is appropriate for the task;

[0103] If the model used needs training, use historical data to train the model;

[0104]

[0105] in, represents the loss function, Y represents the label, θ represents the model parameter, θ * are the model parameters after training;

[0106] Use the trained model to make inferences and predictions on new data;

[0107] Y * =f(X;θ * )

[0108] Where f represents the model function, Y * Represents the results predicted by the model.

[0109] Compared with the prior art, the present invention has the following beneficial effects:

[0110] (1) The present invention adopts a hierarchical structure to effectively solve the problems of complex data processing, difficulty in real-time monitoring, and slow system response speed in industrial production operations. The hierarchical structure ensures the smooth transmission and processing of information between different levels, and improves the overall response efficiency and management effectiveness of the system;

[0111] (2) The present invention effectively addresses the problems of a single intelligent entity being unable to cover diverse production processes, inaccurate task allocation, and single data processing methods in the process of industrial production management through dedicated sub-agent technology. Each dedicated sub-agent has in-depth professional knowledge and precise data processing capabilities in its field, thereby ensuring the accuracy and efficiency of task execution and maximizing the optimization of each link of production operation management. BRIEF DESCRIPTION OF THE DRAWINGS

[0112] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0113] Figure 1 It is the overall architecture diagram of the hierarchical structure;

[0114] Figure 2 It is a flow chart of the intelligent management method;

[0115] Figure 3 The figure is a flow chart of the intelligent management method. DETAILED DESCRIPTION

[0116] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0117] Example

[0118] The present invention provides an intelligent management system for industrial production operations based on the collaboration of multiple AI agents. Users input their needs through various terminals. The AI ​​general agent accurately identifies user needs through natural language processing and user behavior analysis, and assigns tasks to corresponding AI sub-agents according to the characteristics of the needs. Each sub-agent focuses on a specific functional area, such as production management, equipment maintenance, energy management, safety monitoring, etc. The sub-agent receives tasks from the general agent layer and processes them in combination with its own domain knowledge and data. Finally, the processing results are summarized and integrated and fed back to the AI ​​general agent, and the general agent finally returns a result to the user.

[0119] like Figure 1 , the system comprising:

[0120] 1. User layer: interact with the system through user terminals (such as computers, mobile phones, tablets, etc.) to input requirements and instructions.

[0121] 2. General agent layer: Design a central AI general agent to accurately identify user needs through natural language processing and user behavior analysis, and assign tasks to corresponding sub-agents based on the characteristics of the needs.

[0122] 3. Sub-agent layer: Build multiple dedicated AI sub-agents, each of which focuses on a specific functional area, such as production management, equipment maintenance, energy management, safety monitoring, etc. The sub-agent receives tasks from the general agent layer and processes them based on its own domain knowledge and data.

[0123] 4. Tool layer: The sub-agent layer calls different tools to complete specific tasks, including:

[0124] Image processing tools: for visual recognition and image analysis;

[0125] Audio processing tools: for speech recognition and audio analysis;

[0126] Text Generation Tool (LLM): used to generate natural language text and reports;

[0127] Retrieval Enhancement Generation Tool (RAG): used to retrieve and generate relevant information from massive data.

[0128] like Figure 2 and Figure 3 The present invention provides an intelligent management method for industrial production operations based on multi-AI intelligent agent collaboration, comprising the following steps:

[0129] Step 1: The user logs in to the terminal device application;

[0130] Step 2: Understanding user intent and needs analysis;

[0131] (1) Application of Natural Language Processing (NLP) Technology: AI general agent uses NLP technology to parse the natural language input by users, including semantic understanding, intent classification, and entity recognition, to accurately capture user needs.

[0132] 1. Tokenization: Split the text input by the user into a sequence of words or subwords. Common tokenization methods include rule-based tokenization and statistical-based tokenization. Use the Maximum Match Method for Chinese tokenization.

[0133] The Maximum Match Method is one of the common Chinese word segmentation algorithms. Its basic idea is to match the longest words in the text from left to right until the entire text is segmented. Here is the basic description and algorithm example of the Maximum Match Method:

[0134] Algorithm description:

[0135] Input: The Chinese text S to be segmented.

[0136] Output: The word segmentation result Seg, that is, the sequence of segmented words.

[0137] Process:

[0138] Start from the left end of the text and use the longest word in the dictionary as the matching target.

[0139] If the match is successful, add the word to the word segmentation result and delete the word from the text.

[0140] If the match is unsuccessful (that is, no matching word can be found in the dictionary), regard the current character as a single word and add it to the word segmentation result.

[0141] Repeat the above steps until the entire text is completely processed.

[0142] Example:

[0143] Suppose there is a simple Chinese text S:

[0144] S = "今天天气真不错"

[0145] The words included in the dictionary are: {"今天", "天气", "真", "不错"}.

[0146] The process of word segmentation according to the Maximum Match Method is as follows:

[0147] Initial state: The text is "今天天气真不错", and the word segmentation result is empty.

[0148] First step: Match the longest word "今天", add it to the word segmentation result, and delete it from the text. The remaining text is "天气真不错".

[0149] Second step: Match the longest word "天气", add it to the word segmentation result, and delete it from the text. The remaining text is "真不错".

[0150] Third step: Match the longest word "真", add it to the word segmentation result, and delete it from the text. The remaining text is "不错".

[0151] Step 4: Match the longest word "not bad", add it to the word segmentation result, and delete it from the text. The remaining text is empty.

[0152] End: The word segmentation result is {"today", "weather", "really", "not bad"}.

[0153] 2. Part-of-Speech Tagging: Determine the part of speech of each word in the context (such as noun, verb, adjective, etc.).

[0154] 3. Named Entity Recognition (NER): Identify entities with specific meanings in the text, such as person names, place names, organization names, etc.

[0155] Conditional Random Field (CRF) is used for named entity recognition:

[0156] CRF model concept: Given a set of input sequences X = (x1, x2,..., x n ), and the corresponding annotation sequence Y = (y1, y2,..., y n ), the goal of the CRF model is to learn the conditional probability P(Y∣X), where Y is the annotation sequence and X is the input sequence.

[0157] CRF calculation formula: The CRF model obtains the optimal annotation sequence Y by maximizing the conditional probability P(Y∣X) * :

[0158] Y * = argmax P(Y∣X)

[0159] where P(Y∣X) can be expressed as:

[0160]

[0161] where λ k is the weight of the feature function f k and Z(X) is the normalization factor to ensure that the sum of probabilities is 1.

[0162] 4. Semantic Parsing: Convert natural language into a structured representation so that the computer can understand and operate. For example, convert a natural language query into a logical form or query language.

[0163] (2) Intention Inference and Context Understanding: Infer the specific intention of the user by analyzing the context and the user's interaction history, and consider possible implicit information.

[0164] 1. Collect input data

[0165] First, the system needs to collect the user's current input, context information, and interaction history. This data includes:

[0166] The text currently entered by the user;

[0167] The user’s previous interaction history;

[0168] User's personal information (such as preferences, interests, etc.).

[0169] 2. Natural Language Processing (NLP)

[0170] Using NLP technology to preprocess and parse user input includes the following steps:

[0171] Tokenization: Split the text input by the user into a sequence of words or subwords;

[0172] Part-of-Speech Tagging: mark the part of speech of each word;

[0173] Named Entity Recognition (NER): Identify entities in text, such as names of people, places, organizations, etc.

[0174] Syntax Parsing: Analyze the syntactic structure of the text and build a dependency tree or syntax tree.

[0175] 3. Contextual understanding

[0176] Understand the user’s intent by analyzing the context of the current conversation and the user’s interaction history, including:

[0177] Context feature extraction: extract context features of the current conversation, such as topics, keywords, emotions, etc.

[0178] User history analysis: Analyze the user's previous interaction records, including previous queries, click behaviors, feedback, etc.;

[0179] Context modeling: Use context modeling techniques (such as RNN, LSTM, Transformer, etc.) to capture the dynamic characteristics of conversations and long-term and short-term dependencies.

[0180] 4. Intent Recognition

[0181] Leveraging machine learning and deep learning models for intent recognition:

[0182] Intent classification model: Train a classification model (such as SVM, decision tree, neural network, etc.) to identify the user's intent. Common models include LSTM, BERT, etc.

[0183] Feature vector representation: user input, context features, and user history are represented as feature vectors;

[0184] Intent prediction: Input the feature vector into the intent classification model and output the predicted intent category.

[0185] 5. Intention Inference

[0186] Combine context and historical data to perform more advanced intent inference:

[0187] Implicit intent mining: Use implicit intent models (such as Bayesian networks, hidden Markov models, etc.) to mine users’ implicit intent;

[0188] Multi-round dialogue processing: Process multi-round dialogues and make dynamic adjustments and inferences by tracking changes in dialogue status and user intent.

[0189] 6. Requirements refinement and clarification

[0190] Convert the inferred user intent into a clear task description to ensure the accuracy and effectiveness of task decomposition:

[0191] Task decomposition: break down complex user intentions into specific subtasks;

[0192] Clarification and confirmation: Interact with users to clarify unclear intentions and ensure accurate understanding;

[0193] Example process:

[0194] Suppose the user inputs: "I want to book a flight from Beijing to Shanghai tomorrow";

[0195] Collect input data:

[0196] Current input: "I want to book a flight from Beijing to Shanghai tomorrow";

[0197] Contextual information: current conversation topic;

[0198] User history: previous searches and purchases;

[0199] Natural Language Processing (NLP):

[0200] Participle: ["I", "want", "book", "a", "tomorrow", "from", "Beijing", "to", "Shanghai", "of", "air ticket"];

[0201] Part-of-speech tagging: ["PRON","VERB","VERB","NUM","NOUN","PREP","NOUN","PREP","NOUN","PART","NOUN"];

[0202] Entity recognition: {"Beijing":"LOC","Shanghai":"LOC","tomorrow":"DATE"};

[0203] Contextual understanding:

[0204] Extract contextual features of the current conversation;

[0205] Analyze the user’s previous booking records;

[0206] Intent Recognition:

[0207] Feature vector representation;

[0208] The intent classification model predicts the intent as "booking a ticket";

[0209] Intent Inference:

[0210] Leverage the user’s previous search and purchase behavior to infer the flight companies and time periods the user may prefer;

[0211] Requirements refinement and clarification:

[0212] The tasks are broken down into: booking tickets, confirming time, and choosing an airline company;

[0213] Confirm specific requirements with users to ensure accurate understanding.

[0214] (3) Requirements refinement and clarification: Convert abstract user intent into clear task descriptions (see the above case) to ensure the accuracy and effectiveness of task decomposition.

[0215] Step 3: Task decomposition and task distribution;

[0216] (1) Decomposition of complex tasks: The AI ​​general agent breaks down large tasks into small tasks based on user needs. Each small task has a clear execution path and goal.

[0217] (2) Task priority setting: Determine the execution priority of each subtask based on the task's urgency, resource requirements, and dependencies;

[0218] (3) Task allocation to sub-agents: The task coordinator assigns each sub-task to a dedicated AI sub-agent, ensuring that each agent is able to handle tasks within its area of ​​expertise.

[0219] Step 4: Each sub-agent performs tasks and provides feedback;

[0220] (1) Real-time task execution monitoring: Each dedicated AI sub-agent starts to execute the assigned task and monitors the execution progress and status of the task in real time;

[0221] (2) Result generation and recording: Each sub-agent generates corresponding execution results according to task requirements, ensuring that the results contain necessary detailed information and analysis reports;

[0222] (3) Task execution feedback: The sub-agent feeds back the execution results to the task coordinator for subsequent result integration and comprehensive analysis.

[0223] Step 5: Results integration and feedback

[0224] (1) Data aggregation and analysis: The task coordinator receives and integrates the execution results of each sub-agent, and conducts comprehensive analysis and processing of the results through data aggregation methods.

[0225] 1. Data Aggregation

[0226] Data aggregation refers to integrating the execution results from different sub-agents for comprehensive analysis and decision-making. Data aggregation includes the following steps:

[0227] a. Data Collection

[0228] Collecting execution results from each sub-agent, which may include data in various forms, such as structured data, semi-structured data, and unstructured data;

[0229] {R1,R2,...,R n}

[0230] Among them, R i represents the execution result of the i-th sub-agent;

[0231] b. Data cleaning

[0232] Clean and preprocess the collected data to remove redundancy and noise and ensure data quality;

[0233] D clean =Clean(D)

[0234] Where D represents the original data set, D clean represents the cleaned data set;

[0235] c. Data conversion and standardization

[0236] Convert data in different formats into a unified format and perform standardization;

[0237] D norm =Normalize(D clean )

[0238] Among them, D norm Represents the standardized data set;

[0239] d. Data merging

[0240] Combine the results of each sub-agent into a comprehensive data set for subsequent analysis and inference;

[0241] D merged =Merge(D norm,1 ,D norm,2 ,...,D norm,n )

[0242] 2. Model Reasoning

[0243] Model reasoning refers to the use of machine learning or deep learning models to reason and make decisions on aggregated data. Model reasoning includes the following steps:

[0244] a. Feature extraction

[0245] Extract feature vectors from the aggregated dataset for model input;

[0246]

[0247] in, represents the feature extraction function, X represents the feature matrix;

[0248] b. Model selection

[0249] Choose a machine learning or deep learning model that is appropriate for the task, such as decision trees, support vector machines, random forests, neural networks, etc.

[0250] c. Model training

[0251] If the model used needs training, use historical data to train the model;

[0252]

[0253] in, represents the loss function, Y represents the label, and θ represents the model parameter;

[0254] d. Model Reasoning

[0255] Use the trained model to infer and predict new data; Y * =f(X;θ * )

[0256] Where f represents the model function, Y * Represents the results predicted by the model;

[0257] (2) User or system feedback: The AI ​​general agent will feed back the comprehensive final results to the user or system to complete the comprehensive processing and response of task requirements, ensuring the user or system’s satisfaction and effect evaluation of task execution.

[0258] Those skilled in the art know that, in addition to implementing the system, device and its various modules provided by the present invention in a purely computer-readable program code, it is entirely possible to implement the same program in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers and embedded microcontrollers by logically programming the method steps. Therefore, the system, device and its various modules provided by the present invention can be considered as a hardware component, and the modules included therein for implementing various programs can also be considered as structures within the hardware component; the modules for implementing various functions can also be considered as both software programs for implementing the method and structures within the hardware component.

[0259] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. An intelligent management method for industrial production operations based on multi-AI intelligent agent collaboration, characterized in that: include: Step 1: The user logs in to the terminal device application and enters natural language; Step 2: Use natural language processing (NLP) technology to parse the natural language input by the user, including semantic understanding, intent classification, and entity recognition. By analyzing the context and the user's interaction history, the user's specific intent is inferred and the abstract user intent is converted into a clear task description. Step 3: Task decomposition and task distribution; Step 4: Each sub-agent performs tasks and provides feedback; Step 5: Receive and integrate the execution results of each sub-agent through the task coordinator, conduct comprehensive analysis and processing on the results through data aggregation methods, and feed back the comprehensive final results to the user or system to complete the comprehensive processing and response of the task requirements.

2. The method for intelligent management of industrial production operations based on multi-AI agent collaboration according to claim 1 is characterized in that: The step 2 comprises: Use natural language processing (NLP) technology to parse the natural language input by the user, specifically: Word segmentation: split the text input by the user into a sequence of words or subwords. Word segmentation methods include rule-based word segmentation and statistical-based word segmentation. The maximum matching method is used for Chinese word segmentation. Part-of-speech tagging: determining the part of speech of each word in context; Entity recognition: Identify entities with specific meanings in text, including names of people, places, and organizations; Semantic understanding: converting natural language into structured representations; By analyzing the context and the user's interaction history, the user's specific intentions are inferred, specifically: Collect the user's current input, context information and interaction history, including: the text currently input by the user, the user's previous interaction records and the user's personal information; Use NLP technology to pre-process and parse user input, including: word segmentation, part-of-speech tagging, named entity recognition and syntactic parsing; Extract the contextual features of the current conversation, including topics, keywords, and emotions; analyze the user's previous interaction records, including previous queries, click behaviors, and feedback; use context modeling technology to capture the dynamic characteristics of the conversation and long-term and short-term dependencies; Train an intent classification model to identify the user's intent, representing the user input, context features, and user history as a feature vector; input the feature vector to the intent classification model, and output the predicted intent category; Combine context and historical data, and use implicit intent models to mine users’ implicit intents; process multiple rounds of conversations, and make dynamic adjustments and inferences by tracking changes in conversation status and user intents; Transform inferred user intent into clear task descriptions, break down complex user intent into specific subtasks; interact with users to clarify ambiguous intent.

3. The method for intelligent management of industrial production operations based on multi-AI agent collaboration according to claim 1 is characterized in that: The step 3 comprises: Break down large tasks into small tasks based on user needs, and each small task has a clear execution path and goal; Determine the execution priority of each subtask based on the task's urgency, resource requirements, and dependencies; Each subtask is assigned to the AI ​​sub-agent in the corresponding professional field through the task coordinator.

4. The method for intelligent management of industrial production operations based on multi-AI agent collaboration according to claim 1 is characterized in that: The step 4 comprises: Each sub-agent starts to execute the assigned task and monitors the execution progress and status of the task in real time; Each sub-agent generates corresponding execution results according to task requirements; The sub-agents feed back the execution results to the task coordinator for result integration and comprehensive analysis.

5. The method for intelligent management of industrial production operations based on multi-AI agent collaboration according to claim 1 is characterized in that: The step 5 comprises: Data aggregation: Integrate the execution results from different sub-agents for comprehensive analysis and decision-making; The data aggregation process is as follows: Collect execution results from each sub-agent, including structured data, semi-structured data, and unstructured data; {R1,R2,...,R n } Among them, R i represents the execution result of the ith sub-agent; n is the number of sub-agents; The Excel function Clean() is used to clean and preprocess the collected data to remove redundancy and noise and ensure data quality; D clean =Clean(D) Where D represents the original data set, D clean represents the cleaned data set; The normalization function Normalize() is used to convert data in different formats into a unified format and perform standardization. D norm =Normalize(D clean ) Among them, D norm Represents the standardized data set; Use the data merge() function to merge the results of each sub-agent into an aggregate data set D merged , for subsequent analysis and reasoning; D merged =Merge(D norm,1 ,D norm,2 ,...,D norm,n ) Model reasoning: Use machine learning or deep learning models to reason and make decisions on aggregated data. The model reasoning process is: Extract feature vectors from the aggregated dataset for model input; in, represents the feature extraction function, X represents the feature matrix; Choose a machine learning or deep learning model that is appropriate for the task; If the model used needs training, use historical data to train the model; in, represents the loss function, Y represents the label, θ represents the model parameter, θ * are the model parameters after training; Use the trained model to make inferences and predictions on new data; Y * =f(X;θ * ) Where f represents the model function, Y * Represents the results predicted by the model.

6. An industrial production operation intelligent management system based on multi-AI intelligent agent collaboration, characterized in that: include: Module M1: enables the user to log in to the terminal device application and input natural language; Module M2: Use natural language processing (NLP) technology to parse the natural language input by users, including semantic understanding, intent classification, and entity recognition. By analyzing the context and user interaction history, the user's specific intent is inferred and the abstract user intent is converted into a clear task description. Module M3: Task decomposition and task distribution; Module M4: each sub-agent performs tasks and provides feedback; Module M5: Receive and integrate the execution results of each sub-agent through the task coordinator, conduct comprehensive analysis and processing of the results through data aggregation methods, and feed back the comprehensive final results to the user or system to complete the comprehensive processing and response of task requirements.

7. The industrial production operation intelligent management system based on multi-AI intelligent agent collaboration according to claim 6 is characterized in that: The module M2 comprises: Use natural language processing (NLP) technology to parse the natural language input by the user, specifically: Word segmentation: split the text input by the user into a sequence of words or subwords. Word segmentation methods include rule-based word segmentation and statistical-based word segmentation. The maximum matching method is used for Chinese word segmentation. Part-of-speech tagging: determining the part of speech of each word in context; Entity recognition: Identify entities with specific meanings in text, including names of people, places, and organizations; Semantic understanding: converting natural language into structured representations; By analyzing the context and the user's interaction history, the user's specific intentions are inferred, specifically: Collect the user's current input, context information and interaction history, including: the text currently input by the user, the user's previous interaction records and the user's personal information; Use NLP technology to pre-process and parse user input, including: word segmentation, part-of-speech tagging, named entity recognition and syntactic parsing; Extract the contextual features of the current conversation, including topics, keywords, and emotions; analyze the user's previous interaction records, including previous queries, click behaviors, and feedback; use context modeling technology to capture the dynamic characteristics of the conversation and long-term and short-term dependencies; Train an intent classification model to identify the user's intent, representing the user input, context features, and user history as a feature vector; input the feature vector to the intent classification model, and output the predicted intent category; Combine context and historical data, and use implicit intent models to mine users’ implicit intents; process multiple rounds of conversations, and make dynamic adjustments and inferences by tracking changes in conversation status and user intents; Transform inferred user intent into clear task descriptions, break down complex user intent into specific subtasks; interact with users to clarify ambiguous intent.

8. The industrial production operation intelligent management system based on multi-AI intelligent agent collaboration according to claim 6 is characterized in that: The module M3 comprises: Break down large tasks into small tasks based on user needs, and each small task has a clear execution path and goal; Determine the execution priority of each subtask based on the task's urgency, resource requirements, and dependencies; Each subtask is assigned to the AI ​​sub-agent in the corresponding professional field through the task coordinator.

9. The industrial production operation intelligent management system based on multi-AI intelligent agent collaboration according to claim 6 is characterized in that: The module M4 comprises: Each sub-agent starts to execute the assigned task and monitors the execution progress and status of the task in real time; Each sub-agent generates corresponding execution results according to task requirements; The sub-agents feed back the execution results to the task coordinator for result integration and comprehensive analysis.

10. The industrial production operation intelligent management system based on multi-AI intelligent agent collaboration according to claim 6 is characterized in that: The module M5 comprises: Data aggregation: Integrate the execution results from different sub-agents for comprehensive analysis and decision-making; The data aggregation process is as follows: Collect execution results from each sub-agent, including structured data, semi-structured data, and unstructured data; {R1,R2,...,R n } Among them, R i represents the execution result of the ith sub-agent; n is the number of sub-agents; The Excel function Clean() is used to clean and preprocess the collected data to remove redundancy and noise and ensure data quality; D clean =Clean(D) Where D represents the original data set, D clean represents the cleaned data set; The normalization function Normalize() is used to convert data in different formats into a unified format and perform standardization. D norm =Normalize(D clean ) Among them, D norm Represents the standardized data set; Use the data merge() function to merge the results of each sub-agent into an aggregate data set D merged , for subsequent analysis and reasoning; D merged =Merge(D norm,1 ,D norm,2 ,...,D norm,n ) Model reasoning: Use machine learning or deep learning models to reason and make decisions on aggregated data. The model reasoning process is: Extract feature vectors from the aggregated dataset for model input; in, represents the feature extraction function, X represents the feature matrix; Choose a machine learning or deep learning model that is appropriate for the task; If the model used needs training, use historical data to train the model; in, represents the loss function, Y represents the label, θ represents the model parameter, θ * are the model parameters after training; Use the trained model to make inferences and predictions on new data; Y * =f(X;θ * ) Where f represents the model function, Y * Represents the results predicted by the model.

Citation Information

Cited By

  • Education task processing method and device based on multi-agent dialogue driving

    CN119829290A

  • Intelligent dialogue system for time sequence memory management and task distribution

    CN120086338A

  • An intelligent dialogue system for temporal memory management and task distribution

    CN120086338B

  • Intelligent reasoning method and system based on large language model

    CN120106215A

  • Data processing method, data processing system and computing device

    WO2026118534A1