Equipment abnormal information collection method and system based on large model agent

Abstract

The invention provides an equipment exception information collection method and system based on a large model agent, and the method comprises the steps: firstly, building a form file according to the equipment exception information needing to be collected, and configuring the large model agent through a cue word; the cue word comprises a large model agent role as a factory equipment exception information collector and a function of collecting equipment exception information through dialogues, calling an insertion form interface API to fill the information into a pre-constructed form after information collection is completed, and equipment exception information to be collected. And then, in response to a trigger operation of the user on the current session, obtaining a dialogue handle, calling the large model agent to perform a plurality of rounds of dialogues with the user, and collecting equipment abnormal information in the dialogue process. And finally, according to decision information of information collection completion made by the large model agent, calling an insertion form interface API to fill the dialogue handle and the collected information into a pre-constructed form.

Description

Technical Field

[0001] The present invention relates to the technical field of information collection, and in particular to a method and system for collecting equipment abnormality information based on a large model intelligent agent. Background Art

[0002] With the improvement of industrial automation, the complexity of chip processing equipment in semiconductor manufacturing plants continues to increase, resulting in a large amount of information related to equipment abnormalities in the factory. The information is scattered and difficult to quickly precipitate into a structured knowledge base to provide knowledge support for equipment abnormality cause analysis and monitoring parameter recommendation.

[0003] In order to achieve the systematic collection and integration of scattered information, technicians in this field have adopted the following methods: 1. Manually integrate the scattered equipment abnormality information in the technical manuals of each device, and manually record and organize the scattered equipment abnormality knowledge and cases through interviews, questionnaires and knowledge sharing meetings. This information collection method is time-consuming and labor-intensive, and is prone to information errors, incomplete information, untimely information updates, and insufficient structuring.

[0004] 2. Use natural language processing technology to extract equipment abnormality information from documents, reports, emails and other texts. Natural language processing technology can automatically extract key information from a large amount of text and data, and preliminarily organize and summarize this information. However, this method can only be used for equipment abnormality information that has been recorded in text, and cannot integrate manual experience. It still faces the problems of incomplete information, untimely information updates, and insufficient structuring. Summary of the invention

[0005] In response to the above problems, the present invention provides a conversational information collection method and system based on a large model intelligent agent.

[0006] According to a first aspect, the present invention provides a conversational information collection method based on a large model agent, comprising the following steps: In response to the user's trigger operation on the current session, call the conversation handle creation API to obtain the conversation handle ID; The large model agent is called to conduct several rounds of dialogue with the user, where the first round of dialogue includes: receiving a first input from a user; The conversation handle ID, the first input and the pre-configured prompt words are input into the large model agent to obtain the first output of the large model agent; the prompt words include a role description: the role of the large model agent is a factory equipment abnormality information collector, a functional description: collect information through conversation interaction, call the insert form interface API after the information collection is completed to fill the information into a pre-built form, and the equipment abnormality information to be collected; Determine a first response based on the first output, and present the response to the user; According to the decision information made by the large model agent after the information collection is completed, the insert form interface API is called to fill the dialogue handle and the collected information into the pre-built form.

[0007] In some embodiments, the plurality of rounds of dialogues are multiple rounds of dialogues, wherein any round of dialogue other than the first round of dialogue includes: Receive the second input from the user in this round of dialogue; The second input is processed using the large model agent to obtain a second output.

[0008] In some embodiments, the device abnormality information includes the device number, cavity number, abnormality description, abnormality occurrence time, abnormality impact parameter name, abnormality impact parameter SVID, and the device location where the abnormality occurred.

[0009] In some embodiments, the information to be collected in the prompt word also includes user information; the user information includes the user's name and phone number.

[0010] In some embodiments, the prompt word also includes a summary information description: After the device exception information is collected, it is necessary to summarize the severity level of the exception based on the collected device exception information, and summarize the occurrence process of the exception in one sentence, and call the insert form interface API to fill in the severity level and summary into the form.

[0011] After the large model intelligent agent makes a decision on the completion of information collection, the method also includes: the large model intelligent agent summarizes the severity level of the anomaly based on the collected equipment anomaly information, summarizes the occurrence process of the anomaly in one sentence, and calls the insert form interface API to fill in the severity level and summary into the form.

[0012] In some embodiments, the prompt word also includes a description of refusal to answer information: the large model agent needs to refuse to answer when the user asks questions outside the field of device abnormalities.

[0013] In some embodiments, before calling the large model agent to communicate with the user, the method also includes: creating a form file, the header fields of the form file correspond to the information to be filled in; generating a data insertion interface API corresponding to the header fields, and connecting it to the large model agent.

[0014] In some embodiments, the triggering operation of the user on the current session specifically includes: the user clicking on the session logo, or the user inputting information based on the session interface; in response to the triggering operation of the user on the current session, the method further includes: sending an opening greeting to the user according to the triggering operation.

[0015] In some embodiments, the dialog interaction is implemented through a loop dialog interface.

[0016] According to a second aspect, the present invention provides a conversational information collection system based on a large model agent, comprising: The startup module is used to call the conversation handle creation API to obtain the conversation handle ID after the user performs a trigger operation on the current conversation.

[0017] The dialogue module is used to call the large model agent to conduct several rounds of dialogue with the user, where the first round of dialogue includes: receiving a first input from a user; The conversation handle ID, the first input and the pre-configured prompt words are input into the large model agent to obtain the first output of the large model agent; the prompt words include a role description: the role of the large model agent is a factory equipment abnormality information collector, a functional description: collect information through conversation interaction, call the insert form interface API after the information collection is completed to fill the information into a pre-built form, and the equipment abnormality information to be collected; A first reply is determined based on the first output and presented to the user.

[0018] The input module is used to call the insert form interface API to fill the dialogue handle and the collected information into a pre-built form according to the decision information completed by the information collection made by the large model intelligent agent.

[0019] The conversational information collection method and system provided by the present invention can realize the automatic and comprehensive collection and organization of abnormal information of factory equipment, improve the efficiency of information collection, and avoid the extra time spent on manually collecting and summarizing information. The method and system collect information through conversational interaction and can integrate scattered manual experience. In addition, the large model can autonomously understand and integrate the description of equipment abnormal information input by the user, and remind the user to supplement it through dialogue when the information is insufficient, which not only ensures the integrity and structured degree of the information, but also brings a better experience to the user during the information collection process. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 A flow chart of a method for collecting abnormal information of equipment provided by the present invention; Figure 2A flow chart of the preparation phase of a method for collecting abnormal information of equipment provided by the present invention Figure 3 This is a structural diagram of a device abnormality information collection system provided by the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0023] Large Language Model Agents (LLM Agents) is an artificial intelligence system that uses a large language model as its core computing engine, has natural language understanding and generation capabilities, can conduct intelligent dialogues, generate summaries, reason, and can perform a certain degree of autonomous behavior to complete tasks or make decisions based on the functions granted during the design phase. Using the reasoning ability of large language models, agents can manage relatively independent workflows: analyzing goals, project planning, execution, reviewing past work, and iterative refinement.

[0024] In addition, the large language model agent can use a tool kit (such as computing tools, API interfaces, search engines, etc.) to complete the assigned tasks. Using these tools, the large model agent can generate customized text, reports, and pictures, such as emails, reports, and marketing materials.

[0025] The emergence of large-model intelligent agents provides a new idea for the automated collection and organization of information. Based on the large-model intelligent agent, the present invention collects scattered equipment anomaly information through intelligent dialogue, uses the powerful context understanding ability and domain knowledge reserve of the large-model intelligent agent to analyze and summarize the scattered equipment anomaly information, and finally allows the large-model intelligent agent to autonomously call the form interface API, directly fill the collected equipment anomaly information into the form, complete the collection and organization of scattered equipment anomaly knowledge, and form a systematic and structured equipment anomaly knowledge base.

[0026] The specific implementation process of the conversational information collection method provided by the present invention is as follows: Figure 1 As shown, it is divided into the preparation stage before the dialogue begins and the dialogue stage.

[0027] First, we will introduce the preparation stage before the dialogue begins. A specific implementation process of this stage is as follows: Figure 2 As shown, it includes creating form files, configuring large model agents and generating user interfaces.

[0028] 1. Create a form file The form file is used to store the collected device abnormality information and can be an Excel worksheet.

[0029] First, the header field of the form file needs to be designed. The header field corresponds to the information to be filled in, including the device abnormality information to be collected, the user information to be collected, and the summary information obtained by the autonomous analysis of the large model intelligent agent, etc., which can be defined according to the needs.

[0030] The device abnormality information includes the device number, cavity number, abnormality description, abnormality occurrence time, abnormality impact parameter name, abnormality impact parameter SVID, device location where the abnormality occurred, etc. The user information includes the user's name and phone number, etc.

[0031] Next, you need to set appropriate form access permissions (such as read and write permissions), generate a data insertion interface API containing header fields, and generate the corresponding API Key for access authorization.

[0032] Then, insert the data into the interface API to access the large model agent.

[0033] Specifically, you can create a dialogue client thread handle responsible for interacting with the API, write the API Key and the API's Base_URL as tool call parameters into the dialogue client thread handle, and define the calling method of the interface API in the dialogue client thread handle. For example, when the large model agent makes a decision that information collection is complete, call the data insertion interface API to fill in the current dialogue handle and the collected information into the pre-built form.

[0034] 2. Configure large model agents The large model agent used in the present invention can adopt a domain large model as the basic framework. The domain large model is obtained by fine-tuning the pre-trained large language model through instructions, and the fine-tuning sample is constructed based on the domain information with special terms in the field of equipment operation and maintenance. The domain large model can understand the language patterns, special terms and concepts specific to the field, can solve complex problems in a specific field, and show excellent performance in the field. When collecting equipment abnormality information, the domain large model can better understand the special terms in the equipment abnormality information input by the user in the conversation, bringing better user experience and information collection effect.

[0035] After selecting the big model framework, the big model agent is interacted with by using specific prompts, setting corresponding roles for the big model agent and granting corresponding permissions, so that it can become an agent that can perform tasks autonomously. Prompts include: role description, functional description, information to be collected, summary information description, and refusal information description. Among them, the summary information description refers to the content that the big model agent needs to summarize autonomously; the refusal information description refers to the circumstances under which the big model agent needs to refuse to answer.

[0036] In the present invention, the prompt words include: the role of the large model intelligent body is a factory equipment abnormality information collector, and its function is to collect information through dialogue interaction. After the information collection is completed, the insertion form interface API is called to fill the information into a pre-built form, as well as the equipment abnormality information to be collected.

[0037] In some embodiments, the prompt word also includes a summary information description: after the device abnormality information is collected, it is necessary to summarize the severity level of the abnormality based on the collected device abnormality information, and summarize the occurrence process of the abnormality in one sentence.

[0038] In some embodiments, the prompt word also includes a description of refusal to answer information: the large model agent needs to refuse to answer when the user asks questions outside the field of device abnormalities.

[0039] Exemplarily, the prompt text input to the large language model in the large model agent framework can be: "The role of this assistant is to collect abnormal information of FAB plant equipment. It is good at analyzing various abnormal descriptions of chip processing equipment production process by engineers to form a knowledge base. The equipment abnormality information that needs to be collected includes: equipment number, cavity number, abnormality description, abnormality occurrence time, parameter name affected by the abnormality, equipment part where the abnormality occurred, and abnormality process description. In addition, the severity level of this abnormality needs to be summarized according to the engineer's description. The severity levels are divided into: not serious, general, and serious. After collecting information for the engineer dialogue, it is necessary to ask the worker to The engineer leaves his name and phone number. Finally, summarize a sentence to describe the occurrence process of the anomaly so that the equipment anomaly analysis team can conduct further in-depth analysis. After obtaining this information, this assistant needs to call the data insertion interface API to fill in this information into the form. It needs to provide the function with ten parameters: equipment number, cavity number, anomaly description, anomaly occurrence time, parameter name affected by the anomaly, equipment part where the anomaly occurred, anomaly process description, the engineer's personal name and phone number, and anomaly severity level. For each dialogue question answered by the engineer, he should be asked to leave equipment and personal details as much as possible. Regarding other issues outside this field, it is necessary to clearly tell the consulting engineer: Due to strict information confidentiality regulations, I cannot provide you with consulting services on other issues. " 3. Orchestration of Dialogue Interaction Components like Figure 2As shown in the figure, after the big model is configured, the interface API and the configured big model agent can be arranged into a dialogue interaction component as an information collection tool. After the dialogue interaction component is arranged, a dialogue information collection user interface (User Interface, UI) can also be created to interact with users. Specifically, the dialogue interaction component can be copied to the information collection front-end interface by referencing script statements, and the dialogue logo and interaction dialog box can be automatically generated. After that, the abnormal information collection dialogue logo is added to the system interface commonly used by users (i.e. engineers), so that multiple engineers can quickly and easily trigger dialogue interactions and collect equipment abnormal experience information.

[0040] The above describes the preparation phase before the dialogue begins. The following describes the dialogue phase. The dialogue phase includes the following steps: 1. In response to the user's trigger operation on the current session, call the conversation handle creation API to obtain the conversation handle ID.

[0041] The dialog handle ID is an identifier used to manage and track the dialog status.

[0042] In some embodiments, the triggering operation of the user on the current session specifically includes: the user clicking on the session logo, or the user inputting information based on the session interface.

[0043] In some embodiments, in response to the user's trigger operation on the current session, the large model agent will send an opening greeting to the user. The opening greeting may include a self-introduction of the large model agent and prompt the user to input device abnormality information.

[0044] 2. Call the large model agent to have several rounds of dialogue with the user, and collect device abnormality information during the dialogue.

[0045] The first round of dialogue included: A first input from a user is received.

[0046] The conversation handle ID, the first input and the pre-configured prompt words are input into the large model agent to obtain the first output of the large model agent; the prompt words have been pre-configured in the preparation stage and will not be repeated here.

[0047] A first reply is determined based on the first output and presented to the user.

[0048] In some embodiments, the plurality of rounds of dialogues are multiple rounds of dialogues, wherein any round of dialogue other than the first round of dialogue includes: Receive the second input from the user in this round of dialogue.

[0049] The second input is processed using the large model agent to obtain a second output.

[0050] In this way, multiple rounds of dialogue are continued until the information collection is completed.

[0051] In some embodiments, the dialogue interaction is implemented through a cyclic dialogue interface. The cyclic dialogue interface is used to ensure that multiple rounds of dialogue can continue until a termination condition is reached, and the termination condition includes that information collection is completed.

[0052] 3. Based on the decision information collected by the large model agent, call the insert form interface API to fill the conversation handle and the collected information into the pre-built form.

[0053] The above is the device abnormality information collection method provided by the present invention. It can be seen that the method has the following beneficial effects: 1. It can realize the automatic collection and organization of abnormal information of factory equipment, improve the efficiency of information collection, and avoid the extra time spent on manual collection and summary of information.

[0054] 2. This method collects information through dialogue interaction and can collect and integrate scattered manual experience.

[0055] 3. The large model intelligent agent can autonomously understand the scattered equipment abnormality information input by the user, automatically summarize the key information based on the user input, and fill the information into the form in strips to form a structured knowledge base.

[0056] 4. The large-model intelligent body can instantly judge the completeness of information and remind users to supplement it through dialogue when the information is insufficient to ensure that the information is collected completely.

[0057] On the other hand, the present invention also provides a device abnormality information collection system 300, the system schematic diagram is shown in FIG. Figure 3 As shown, including: The start module 301 is used to call the conversation handle creation API to obtain the conversation handle ID after the user performs a trigger operation on the current conversation.

[0058] The dialogue module 302 is used to call the large model agent to conduct several rounds of dialogue with the user, wherein the first round of dialogue includes: receiving a first input from a user; The conversation handle ID, the first input and the pre-configured prompt words are input into the large model agent to obtain the first output of the large model agent; the prompt words include a role description: the role of the large model agent is a factory equipment abnormality information collector, a functional description: collect information through conversation interaction, call the insert form interface API after the information collection is completed to fill the information into a pre-built form, and the equipment abnormality information to be collected; Determine a first response based on the first output, and present the response to the user; The input module 303 calls the insert form interface API to fill the conversation handle and the collected information into a pre-built form according to the decision information made by the large model agent after the information collection is completed.

[0059] It should be noted that the above system can execute the above-mentioned device abnormality information collection method. For the functions of each part, please refer to the above-mentioned introduction to the method and will not be repeated here.

[0060] In the description of the embodiments of the present application, words such as "exemplary", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "for example" or "for example" is intended to present related concepts in a concrete way.

[0061] In the description of the embodiments of the present application, the term "and / or" is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, B exists alone, and A and B exist at the same time. In addition, unless otherwise specified, the term "plurality" means two or more.

[0062] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprises", "has" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0063] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solution of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for collecting device abnormality information, characterized in that: The following steps are involved: In response to the user's trigger operation on the current session, call the conversation handle creation API to obtain the conversation handle ID; The large model agent is called to conduct several rounds of dialogue with the user, where the first round of dialogue includes: receiving a first input from a user; The conversation handle ID, the first input and the pre-configured prompt words are input into the large model agent to obtain the first output of the large model agent; the prompt words include a role description: the role of the large model agent is a factory equipment abnormality information collector, a functional description: collect information through conversation interaction, call the insert form interface API after the information collection is completed to fill the information into a pre-built form, and the equipment abnormality information to be collected; Determine a first response based on the first output, and present the response to the user; According to the decision information of the information collection completed by the large model intelligent agent, the insert form interface API is called to fill the dialogue handle and the collected information into the pre-built form.

2. The method according to claim 1, characterized in that: The multiple rounds of dialogues are multiple rounds of dialogues, wherein any round of dialogue other than the first round of dialogue includes: Receive the second input from the user in this round of dialogue; The second input is processed using the large model agent to obtain a second output.

3. The method according to claim 1, characterized in that: The device abnormality information includes the device number, cavity number, abnormality description, abnormality occurrence time, abnormality impact parameter name, abnormality impact parameter SVID, and the device location where the abnormality occurred.

4. The method according to claim 1, characterized in that: The information to be collected in the prompt words also includes user information; the user information includes the user's name and telephone number.

5. The method according to claim 1, characterized in that The prompt words also include a summary information description: after the large model intelligent agent equipment abnormality information is collected, it is necessary to summarize the severity level of the abnormality according to the collected equipment abnormality information, and summarize the occurrence process of the abnormality in one sentence, and call the insert form interface API to fill in the severity level and summary into the form; After the large model intelligent agent makes a decision on the completion of information collection, the method also includes: the large model intelligent agent summarizes the severity level of the exception based on the collected equipment exception information, summarizes the occurrence process of the exception in one sentence, and calls the insert form interface API to fill in the severity level and summary into the form.

6. The method according to claim 1, characterized in that The prompt words also include a description of refusal to answer information: the large model intelligent agent needs to refuse to answer when the user asks questions outside the field of device abnormalities.

7. The method according to claim 1, characterized in that Before calling the large model agent to communicate with the user, the method also includes: creating a form file, the header fields of the form file correspond to the information to be filled in; generating a data insertion interface API corresponding to the header fields, and connecting it to the large model agent.

8. The method according to claim 1, characterized in that: The triggering operation of the user on the current session specifically includes: the user clicking on the session logo, or the user inputting information based on the session interface; in response to the triggering operation of the user on the current session, the method further includes: sending an opening greeting to the user according to the triggering operation.

9. The method according to claim 1, characterized in that: The dialogue interaction is implemented through a circular dialogue interface.

10. A device abnormality information collection system, characterized in that: include: The startup module is used to call the conversation handle creation API to obtain the conversation handle ID after the user performs a trigger operation on the current conversation; The dialogue module is used to call the large model agent to conduct several rounds of dialogue with the user, where the first round of dialogue includes: receiving a first input from a user; The conversation handle ID, the first input and the pre-configured prompt words are input into the large model agent to obtain the first output of the large model agent; the prompt words include a role description: the role of the large model agent is a factory equipment abnormality information collector, a functional description: collect information through conversation interaction, call the insert form interface API after the information collection is completed to fill the information into a pre-built form, and the equipment abnormality information to be collected; Determine a first response based on the first output, and present the response to the user; The input module calls the insert form interface API to fill the conversation handle and the collected information into a pre-built form according to the decision information completed by the information collection made by the large model intelligent agent.

Images