CN-122019575-A - Working procedure splitting method and device of engineering drawing, product and medium

CN122019575ACN 122019575 ACN122019575 ACN 122019575ACN-122019575-A

Abstract

A process splitting method for engineering drawings comprises the steps of receiving engineering drawings, calling a multi-mode large model to split the engineering drawings to generate original process execution data, analyzing to obtain query sentences, searching a plurality of sub-knowledge bases in a hierarchical knowledge base until response information is obtained through searching, optimizing to generate optimized process execution data, traversing the rest sub-knowledge bases to obtain comparison information, judging whether the response information and the comparison information conflict, and outputting the conflict between the response information and the comparison information along with the optimized process execution data if the judgment result is yes. By adopting the scheme, the illusion problem of the multi-mode large model can be effectively restrained, so that the finally output process execution data has higher accuracy and reliability, and can adapt to the differentiated requirements of different factories and clients.

Inventors

QIU ZIXUAN

Assignees

上海黑湖网络科技有限公司
上海黑湖科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (15)

1. The process splitting method of the engineering drawing is characterized by comprising the following steps of: Receiving an engineering drawing of a product to be processed, and calling a multi-mode large model trained in advance to split the working procedure of the engineering drawing to generate original working procedure execution data, wherein the original working procedure execution data comprises a plurality of working procedure steps; Invoking the multi-modal large model to analyze the original process execution data to obtain a query statement, wherein the query statement is generated by the multi-modal large model aiming at least one process step in the plurality of process steps; sequentially retrieving a plurality of sub-knowledge bases in the hierarchical knowledge base according to the order of the priority from high to low based on the query statement until the response information is retrieved, wherein the plurality of sub-knowledge bases have different priorities; Invoking the multi-mode large model, optimizing the original process execution data based on the retrieved response information, and generating optimized process execution data; Traversing and searching other sub-knowledge bases to obtain comparison information, and judging whether the response information conflicts with the comparison information, wherein the other sub-knowledge bases refer to all sub-knowledge bases except the response information; And if the judgment result is yes, outputting the conflict part between the response information and the comparison information together with the optimized process execution data.
2. The process splitting method of engineering drawings according to claim 1, wherein the plurality of sub-knowledge bases comprises: A customer preference sub-knowledge base records preference information of customers entrusting the products to be processed; The factory information sub-knowledge base records factory information of the product to be processed currently; the domain sub-knowledge base records general information of the product category to which the product to be processed belongs; The priority of the client preference sub-knowledge base is higher than that of the factory information sub-knowledge base, and the priority of the factory information sub-knowledge base is higher than that of the field sub-knowledge base.
3. The process splitting method of engineering drawings according to claim 2, wherein the preference information of the customer is stored in the customer preference sub-knowledge base with the identification of the customer or order as an index.
4. The process splitting method of engineering drawing according to claim 2, wherein retrieving response information in the hierarchical knowledge base according to priorities of the plurality of sub-knowledge bases comprises: sequentially inquiring the plurality of sub-knowledge bases according to the priority from high to low until the response information is inquired; And acquiring the response information.
5. The process splitting method of engineering drawing according to claim 4, further comprising: If the response information conflicts with the comparison information, the multi-mode big model is called to obtain conflict suggestion information, wherein the conflict suggestion information comprises description of conflict and suggestion of alternative scheme; And sending a prompt message, wherein the prompt message comprises the conflict suggestion information.
6. The process splitting method of engineering drawing according to claim 5, further comprising, after the sending of the prompt message: Receiving a user decision result input by a user aiming at the prompt message; and if the user decision result comprises an applicable target sub-knowledge base, re-optimizing the original procedure execution data based on comparison information corresponding to the target sub-knowledge base, wherein the target sub-knowledge base is selected from the rest sub-knowledge bases.
7. The process splitting method of engineering drawing according to claim 1, wherein the process step includes a plurality of elements, and the calling the multi-modal large model analyzes the original process execution data to obtain a query sentence includes: judging whether each working procedure is complete or not according to the elements; if the judgment result is negative, carrying out semantic analysis on the step of the working procedure, and extracting missing elements; and calling the multi-mode large model according to the missing elements to obtain the query statement.
8. The process splitting method of engineering drawing according to claim 1, further comprising: Acquiring update information of the hierarchical knowledge base, wherein the update information is obtained by extracting data based on the optimized procedure execution; invoking the multi-mode large model to audit the updated information; And storing the updated information which passes the auditing into a corresponding sub-knowledge base according to a predefined hierarchical rule.
9. The process splitting method of engineering drawing according to claim 1, further comprising: The following steps are repeatedly executed until a preset iteration stop condition is met: Invoking the multi-mode large model to analyze the process execution data after the previous optimization to obtain a query statement, wherein the query statement is generated by the multi-mode large model aiming at least one process step in the plurality of process steps; Based on the query statement, sequentially retrieving a plurality of sub-knowledge bases in the hierarchical knowledge base according to the order of priority from high to low until response information is retrieved; And calling the multi-mode large model, and re-optimizing the optimized process execution data based on the retrieved response information.
10. The process separation method of engineering drawings according to claim 9, wherein the preset iteration stop condition includes that the number of repeated execution reaches a preset iteration number.
11. The process splitting method of engineering drawing according to claim 9, wherein the preset iteration stop condition includes that the query statement is not generated after the multi-mode big model is called to analyze the process execution data after the previous optimization.
12. The process splitting method of engineering drawing according to claim 1, further comprising: comparing the optimized process execution data with the original process execution data, and identifying the changed process steps; Generating an identifier for the changed working procedure step, wherein the identifier is used for indicating the working procedure step, the corresponding query statement and the response information in the original working procedure execution data.
13. The utility model provides a process split device of engineering drawing which characterized in that includes: The drawing receiving and initial dismantling unit is used for receiving an engineering drawing of a product to be processed, calling a multi-mode large model trained in advance to split working procedures of the engineering drawing, and generating original working procedure execution data, wherein the original working procedure execution data comprises a plurality of working procedure steps; The query statement generation unit is used for calling the multi-mode large model to analyze the original process execution data to obtain a query statement, and the query statement is generated by the multi-mode large model aiming at least one process step in the plurality of process steps; the hierarchical retrieval unit is used for sequentially retrieving a plurality of sub-knowledge bases in the hierarchical knowledge base from high to low according to the priority based on the query statement until the response information is retrieved, wherein the sub-knowledge bases have different priorities; the procedure optimizing unit is used for calling the multi-mode large model, optimizing the original procedure execution data based on the retrieved response information and generating optimized procedure execution data; the conflict processing unit is used for traversing and searching the rest sub-knowledge bases to obtain comparison information and judging whether the response information conflicts with the comparison information or not; And the result output unit is used for outputting the conflict part of the response information and the comparison information along with the optimized process execution data if the judgment result is yes.
14. A computer-readable storage medium, the computer-readable storage medium being a non-volatile storage medium or a non-transitory storage medium, on which a computer program is stored, characterized in that the computer program is executed by the process splitting method of the engineering drawing according to any one of claims 1 to 12 when the computer program is run by a computer.
15. A computer program product comprising computer program/instructions, characterized in that the process splitting method of an engineering drawing according to any one of claims 1 to 12 is performed when the computer program/instructions is executed by a computer.

Description

Working procedure splitting method and device of engineering drawing, product and medium Technical Field The invention relates to the technical field of intelligent manufacturing and process design, in particular to a process splitting method and device, a product and a medium of engineering drawings. Background In the field of industrial manufacturing, it is a core task to disassemble engineering drawings of products, such as CAD drawings, into specific processing steps (hereinafter referred to simply as "disassembly"). The traditional manual splitting type has low efficiency and is unfavorable for standardized management. Along with the development of artificial intelligence, the engineering drawing is understood by utilizing a general multi-mode large model in an enterprise information system, and specific processing procedure steps are automatically generated to finish the sheet disassembly, so that the engineering drawing is an important research direction of a person skilled in the art. In the prior art, engineering drawings are input into a general multi-mode large model, and a single-splitting scheme is directly output, so that the technical defect is remarkable, namely, knowledge mastered by the general large model mainly comes from public general data, and depth expertise in the specific manufacturing field is lacked. Therefore, when processing drawings involving complex process logic, special tolerance requirements or multi-equipment coordination, the model is prone to generate "illusions", namely, the process steps of generating elements are missing and the practicality is poor, so that the reliability of planning results cannot meet the industrial production requirements. Those skilled in the art have attempted to introduce search enhancement generation (RAG, RETRIEVAL-Augmented Generation) techniques to reduce the effect of "illusions" on the output results. However, when the scheme processes complex engineering drawings, the search results are fragmented, and complete and reasonable sheet disassembly results are difficult to output. Disclosure of Invention In order to solve the above problems, the present invention at least aims to provide a process splitting method of an engineering drawing, which optimizes the original process execution data by generating a query statement and retrieving a plurality of sub-knowledge bases in a hierarchical knowledge base according to priorities, so as to effectively inhibit the influence of illusion generated by a general multi-mode large model on the splitting result, and make the finally output process execution data have higher integrity and practicability. In a first aspect, the present invention provides a process splitting method for engineering drawings, the method comprising: Receiving an engineering drawing of a product to be processed, and calling a multi-mode large model trained in advance to split the working procedure of the engineering drawing to generate original working procedure execution data, wherein the original working procedure execution data comprises a plurality of working procedure steps; Invoking the multi-modal large model to analyze the original process execution data to obtain a query statement, wherein the query statement is generated by the multi-modal large model aiming at least one process step in the plurality of process steps; sequentially retrieving a plurality of sub-knowledge bases in the hierarchical knowledge base according to the order of the priority from high to low based on the query statement until the response information is retrieved, wherein the plurality of sub-knowledge bases have different priorities; Invoking the multi-mode large model, optimizing the original process execution data based on the retrieved response information, and generating optimized process execution data; Traversing and searching other sub-knowledge bases to obtain comparison information, and judging whether the response information conflicts with the comparison information, wherein the other sub-knowledge bases refer to all sub-knowledge bases except the response information; And if the judgment result is yes, outputting the conflict part between the response information and the comparison information together with the optimized process execution data. Optionally, the plurality of sub-knowledge bases includes: A customer preference sub-knowledge base records preference information of customers entrusting the products to be processed; The factory information sub-knowledge base records factory information of the product to be processed currently; the domain sub-knowledge base records general information of the product category to which the product to be processed belongs; The priority of the client preference sub-knowledge base is higher than that of the factory information sub-knowledge base, and the priority of the factory information sub-knowledge base is higher than that of the field sub-knowledge base. Optionally, the preference inf