Part intelligent precise machining quotation method and system based on large language model

Abstract

The invention discloses an intelligent precise part machining quotation method and system based on a large language model, and belongs to the technical field of intelligent manufacturing and artificial intelligence. Semantic analysis is performed on unstructured manufacturing demand information through the large language model, and key manufacturing process parameters are extracted; if the demand is not complete, a structured process instruction list is output through multiple rounds of clarification inquiry; after the three-dimensional part model is analyzed, machining features are recognized, and when to-be-confirmed features exist, a machining feature list is generated through a man-machine collaborative confirmation mechanism; calculating technical cost according to the structured process instruction list and the processing feature list; non-technical factor information is obtained through a large language model, verification and adjustment are carried out until compliance is achieved, and non-technical cost is calculated; integrating the two types of costs to obtain a total quotation, visually displaying the total quotation and the cost composition, and outputting a quotation sheet after confirmation; the problems that an existing quotation system is difficult in natural language demand analysis, unreliable in feature recognition and lack of an interaction confirmation mechanism are solved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing and artificial intelligence technology, specifically relating to a method and system for intelligent and precise machining and pricing of parts based on a large language model. Background Technology

[0002] In today's increasingly competitive global manufacturing landscape, the ability to quickly and accurately quote prices has become crucial for manufacturing enterprises, especially small and medium-sized enterprises (SMEs), to win orders and enhance their market competitiveness. Quotation is not merely about calculating product costs; it's a comprehensive reflection of a company's manufacturing capabilities and market responsiveness. Traditional manufacturing quotation processes heavily rely on manual experience. Process engineers must repeatedly communicate and interpret vague customer requirements described in natural language (such as "milling grooves here, approximately 5mm deep, with a smooth surface"), and then use 3D CAD (Computer-Aided Design) models to decompose processes, estimate time, and calculate costs. This process often requires cross-departmental collaboration, taking anywhere from hours to days, resulting in low efficiency and susceptibility to quotation errors due to information asymmetry or differences in experience.

[0003] At the technology application level, existing solutions mainly fall into two categories: First, computer-aided quoting (CAQ) systems based on structured data input. While these systems can achieve a certain degree of automation, they cannot effectively understand and process unstructured natural language input, creating a semantic gap between demand communication and cost estimation. Second, integrated environments relying on professional CAD / CAM (Computer-Aided Manufacturing) platforms (such as Siemens NX and PTC Creo). Although these platforms are powerful, their high licensing fees and closed ecosystems exclude resource-constrained small and micro enterprises. Furthermore, all of these systems lack mechanisms for interactive confirmation with users at key decision points (such as when there is uncertainty in the results of processing feature identification), making the quotation process a one-way, "black box" process. Its accuracy depends entirely on the recognition accuracy of the backend algorithm, making it unable to cope with the challenges posed by complex models or ambiguous demands.

[0004] In recent years, with the rapid development of artificial intelligence technology, Large Language Models (LLMs) have demonstrated powerful capabilities in natural language understanding and reasoning, providing a new technological path for intelligent decision-making in manufacturing. Existing research has attempted to introduce semantic technologies (such as ontology and knowledge systems) into the field of cost estimation, performing semantic reasoning through predefined manufacturing rules. However, the capabilities of these methods are limited by pre-built knowledge bases, making it difficult to handle unstructured natural language input in open domains, and lacking dynamic interaction capabilities with users. Meanwhile, in 3D model processing, graph-based feature recognition methods (such as Attributed Adjacency Graphs (AAGs)) have a long history of research, but their accuracy and reliability in recognizing complex features (such as curved surfaces and intersecting features) still need improvement, and no research has yet deeply integrated them with the semantic interaction capabilities of LLMs to build an end-to-end intelligent pricing system.

[0005] Figure 7 The existing quotation system's information selection diagram shows that CNC stands for Computer Numerical Control, and ISO stands for International Organization for Standardization. Manufacturing processes and additional requirements can only be selected from preset options. Clearly, the existing quotation system lacks a natural language-based intelligent human-computer interaction interface and cannot understand the user's ambiguous processing needs. Existing model-driven quotation systems have a unidirectional, non-interactive processing flow, unable to initiate collaborative confirmation with the user at key nodes (such as when feature recognition is uncertain), resulting in quotation accuracy being limited by the algorithm's recognition precision. Existing technical solutions are mostly single-function tools (such as pure feature recognition software or pure cost calculation software), failing to build a complete system integrating "interactive understanding of needs, intelligent model analysis, human-computer collaborative decision-making, and accurate quotation generation."

[0006] Therefore, the manufacturing industry, especially small and micro enterprises, urgently needs a new type of intelligent pricing solution. This solution should be able to utilize LLM technology to understand ambiguous requirements described in natural language, clarify key process parameters through intelligent interaction with users, possess highly reliable processing feature recognition capabilities, and initiate human-machine collaborative confirmation for recognition uncertainties. Ultimately, it should form a complete system integrating "requirement understanding - model parsing - interactive confirmation - cost calculation," providing enterprises with accurate, fast, and reliable pricing services in a low-cost and easily deployable manner. Summary of the Invention

[0007] The purpose of this invention is to overcome the problems of existing quotation systems, such as difficulty in parsing natural language requirements, unreliable feature recognition, and lack of interactive confirmation mechanisms. It proposes a method and system for intelligent and precise machining quotation of parts based on a large language model.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides an intelligent and precise part machining quotation method based on a large language model, comprising: The system acquires unstructured manufacturing requirements information and 3D part models, uses a large language model to perform semantic parsing on the unstructured manufacturing requirements information, and extracts key manufacturing process parameters. It then combines a pre-set manufacturing domain knowledge rule base to determine whether the unstructured manufacturing requirements information is complete. If it is incomplete, the large language model generates a clarifying query and interacts with the user in multiple rounds until the unstructured manufacturing requirements information is complete, and outputs a list of structured process instructions with key manufacturing process parameters. After analyzing the 3D part model, a feature recognition algorithm is used to identify the machining features and determine whether there are any features to be confirmed. If there are no features to be confirmed, the machining feature list is obtained directly. If there are features to be confirmed, a human-machine collaborative confirmation mechanism is executed to obtain the machining feature list. The system calls upon a pre-defined manufacturing domain knowledge rule base and a pre-defined real-time price database. Based on a structured list of process instructions and a list of processing features containing key manufacturing process parameters, it calculates the technical costs. It then interacts with the user through a large language model to obtain information on non-technical factors. Combining this with the manufacturing domain knowledge rule base and actual feasibility, it judges whether the user's needs are reasonable. If they are not reasonable, it prompts the user and negotiates adjustments until the user's needs are reasonable, thus obtaining the non-technical user needs. Based on these non-technical user needs, it calculates the non-technical costs. The total price is obtained by integrating technical and non-technical costs. The total price and cost composition are visualized and displayed. After user confirmation, the price list is generated.

[0009] Furthermore, the input forms of unstructured manufacturing demand information include voice input and text input; The format of the 3D part model is the STEP format, the product model data exchange standard. Key manufacturing process parameters include material type, machining accuracy, surface treatment method, special process requirements, and delivery time; Incomplete information regarding unstructured manufacturing requirements includes missing, ambiguous, or conflicting information.

[0010] Furthermore, when analyzing the 3D part model, a geometric processing kernel based on boundary representation B-Rep is used; The feature recognition algorithm employs an enhanced feature recognition algorithm; Machining features include holes, grooves, cavities, and chamfers.

[0011] Furthermore, if there are features to be confirmed, the human-machine collaborative confirmation mechanism includes: Mark the areas of features to be confirmed in the visualization interface of the 3D part model to obtain the areas to be confirmed; For the area to be confirmed, multiple-choice or question-and-answer confirmation prompts are generated using a large language model to interact with the user and obtain the confirmation result; Based on the confirmation results, the processing feature parameters are corrected to obtain a collaboratively confirmed list of processing features; Output a list of collaboratively confirmed processing features.

[0012] Furthermore, the manufacturing domain knowledge rule base includes cutting parameters, time quotas, and material density; The real-time price database includes material unit prices and outsourcing processing quotations; Technical costs include material costs, processing costs, heat treatment costs, and surface treatment costs; Non-technical factors include delivery location, logistics method, and expected delivery time; Non-technical costs include logistics fees and expedited processing fees.

[0013] Furthermore, the visualization includes cost breakdown charts, and the quotation includes itemized costs, total costs, and delivery dates. The quotation is output in PDF format.

[0014] Secondly, this invention provides an intelligent and precise parts machining quotation system based on a large language model, comprising: The intelligent demand analysis and interactive clarification module is used to acquire unstructured manufacturing demand information and 3D part models. It uses a large language model to perform semantic analysis on the unstructured manufacturing demand information and extract key manufacturing process parameters. Combined with a preset manufacturing domain knowledge rule base, it judges whether the unstructured manufacturing demand information is complete. If it is incomplete, the large language model generates clarification queries and interacts with the user in multiple rounds until the unstructured manufacturing demand information is complete. Finally, it outputs a list of structured process instructions with key manufacturing process parameters. The feature recognition and human-machine collaborative confirmation module is used to analyze the 3D part model, identify the processing features using a feature recognition algorithm, and determine whether there are any features to be confirmed. If there are no features to be confirmed, the processing feature list is obtained directly. If there are features to be confirmed, the human-machine collaborative confirmation mechanism is executed to obtain the processing feature list. The cost tiered calculation and negotiation module is used to call the preset manufacturing domain knowledge rule base and the preset real-time price database. Based on the structured process instruction list and processing feature list with key manufacturing process parameters, it calculates the technical cost. It obtains non-technical factor information through interaction with the user via a large language model. It judges whether the user's demand is reasonable by combining the manufacturing domain knowledge rule base and the actual execution feasibility. If it is unreasonable, it prompts the user demand and negotiates with him to adjust it until the user demand is reasonable. It then obtains the non-technical user demand and calculates the non-technical cost based on the non-technical user demand. The quotation output module is used to integrate technical and non-technical costs to obtain a total quotation, visualize the total quotation and cost composition, and output a quotation sheet after user confirmation.

[0015] Furthermore, it also includes: The user interaction layer is used to provide a 3D model visualization window, voice or text input boxes, intelligent dialogue flow interface, quotation result display panel, receive user input, render models and dialogues, and initiate voice recognition operations. The intelligent service layer is built on a microservice architecture and deployed in the cloud or on a local server to provide services such as large language model interaction, geometry processing, cost calculation, and business logic. The data and knowledge layer is used to store structured data, including material parameters, processing time and equipment rates, using a knowledge rule base in the manufacturing field. By accessing dynamic data, including market material prices and outsourcing quotations, a real-time price database is formed. The project and quotation database stores users' historical projects, model files and quotations. The computing support layer provides the hardware and basic software support required for the operation of the user interaction layer, intelligent service layer, and data and knowledge layer. A voice input device is used to convert input voice into text in real time, and the converted text is transmitted to a large language model for interactive services. A 3D model preview device is used to integrate a part visualization window and highlight specific features; A conversational interactive device used to simulate the interaction format of instant messaging software, presenting a dialogue between a user and an artificial intelligence (AI) assistant.

[0016] Thirdly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a method for intelligent and precise machining quotation of parts based on a large language model.

[0017] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a method for intelligent and precise machining quotation of parts based on a large language model.

[0018] Compared with the prior art, the present invention has the following beneficial technical effects: This invention proposes an intelligent and precise part processing quotation method based on a large language model. This method achieves accurate understanding and efficient parsing of unstructured natural language requirements, avoiding the difficulties in parsing natural language requirements, unreliable feature recognition, and lack of interactive confirmation mechanisms in existing quotation systems. It fundamentally solves the problem of low efficiency in requirement communication: by introducing a large language model as the core of intelligent interaction, the system can directly understand users' vague or colloquial processing requirements and automatically extract key process parameters, completely changing the traditional system's reliance on menu selection and structured input interaction mode. This transforms the previously time-consuming requirement clarification process, which required repeated communication among process engineers, into an automated interaction within minutes or even seconds, greatly improving the efficiency and accuracy of requirement communication and effectively solving the semantic gap problem inherent in traditional methods. By introducing a human-machine collaborative confirmation mechanism, the absolute accuracy of the quotation basis is ensured: to address the uncertainties that may exist in feature recognition, a human-machine interactive confirmation process is proposed, with users making collaborative decisions, thereby integrating human intelligence into the automated process. This completely avoids the risk of traditional one-way processing systems outputting incorrect results due to algorithm limitations. It makes the accuracy of quotations no longer entirely dependent on the recognition rate of the backend algorithm, but is based on the reliable foundation of human-machine collaboration, fundamentally solving the problem of inaccurate quotations caused by false detection and missed detection. Attached Figure Description

[0019] The accompanying drawings described herein are for illustrative purposes only and are not intended to limit the scope of the invention in any way. Furthermore, the shapes and proportions of the components in the drawings are merely illustrative to aid in understanding the invention and do not specifically limit the shapes and proportions of the components of the invention.

[0020] In the attached diagram: Figure 1 This is a flowchart of the intelligent and precise part processing quotation method based on a large language model according to the present invention.

[0021] Figure 2 This is a simplified structural diagram of the intelligent and precise machining quotation system for parts based on a large language model, as described in this invention.

[0022] Figure 3 This is a schematic diagram of an electronic device for the intelligent and precise part processing quotation method based on a large language model according to the present invention.

[0023] Figure 4This is a flowchart of the intelligent quotation method in an embodiment of the present invention.

[0024] Figure 5 This is a diagram illustrating the overall system architecture in an embodiment of the present invention.

[0025] Figure 6 This is a schematic diagram of a quotation sheet in an embodiment of the present invention.

[0026] Figure 7 A diagram illustrating the selection of information from the existing quotation system. Detailed Implementation

[0027] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0028] Example 1 See Figure 1 A method for intelligent and precise machining quotation of parts based on a large language model includes: The system acquires unstructured manufacturing requirements information and a 3D part model. It then uses a large language model to semantically parse the unstructured requirements information, extracting key manufacturing process parameters. Combining this with a pre-defined manufacturing domain knowledge rule base, it determines the completeness of the unstructured requirements information. If incomplete, the large language model generates a clarifying query and engages in multiple rounds of interaction with the user until the requirements are complete, outputting a list of structured process instructions containing key manufacturing process parameters. After parsing the 3D part model, a feature recognition algorithm identifies processing features and determines if any features require confirmation. If no features require confirmation, the system directly obtains a list of processing features. If features require confirmation, a human-machine collaborative confirmation mechanism is executed to obtain the final list. The system generates a processing feature list; it calls upon a pre-defined manufacturing domain knowledge rule base and a pre-defined real-time price database, and calculates the technical cost based on a structured process instruction list with key manufacturing process parameters and a processing feature list; it then interacts with the user through a large language model to obtain non-technical factor information, and judges the rationality of the user's needs by combining the manufacturing domain knowledge rule base and actual execution feasibility. If the user's needs are unreasonable, the system provides prompts and engages in interactive negotiation to adjust them until the user's needs are reasonable, thus obtaining the non-technical user needs. Based on these non-technical user needs, the system calculates the non-technical cost; finally, it integrates the technical and non-technical costs to obtain a total quote, visualizes the total quote and cost composition, and outputs a quote after user confirmation.

[0029] This embodiment uses a large language model to semantically parse unstructured requirements, accurately extracting key process parameters. It then combines this with a manufacturing knowledge rule base to complete the requirements, avoiding information bias. Three-dimensional model feature recognition, combined with human-machine collaborative confirmation, ensures no processing features are missed. This is further linked to a real-time price database to calculate technical costs, and non-technical factors are interactively verified, resulting in a final quote that closely matches the actual scenario. Automated parsing replaces manual information screening, multi-round clarification inquiries shorten the requirement confirmation cycle, and standardized feature recognition and cost calculation processes significantly reduce manual operation time and improve quote response speed. The human-machine collaborative mechanism solves complex problems such as feature confirmation and requirement rationality judgment. Visualized quote display reduces user comprehension costs, and multi-round interaction enhances consensus between supply and demand parties, reducing subsequent negotiations and rework, balancing technical rigor with user experience.

[0030] Example 2 See Figure 2 A parts intelligent and precise machining quotation system based on a large language model includes: The intelligent demand analysis and interactive clarification module is used to acquire unstructured manufacturing demand information and 3D part models. It uses a large language model to perform semantic analysis on the unstructured manufacturing demand information and extract key manufacturing process parameters. Combined with a preset manufacturing domain knowledge rule base, it judges whether the unstructured manufacturing demand information is complete. If it is incomplete, the large language model generates clarification queries and interacts with the user in multiple rounds until the unstructured manufacturing demand information is complete. Finally, it outputs a list of structured process instructions with key manufacturing process parameters. The feature recognition and human-machine collaborative confirmation module is used to analyze the 3D part model, identify the processing features using a feature recognition algorithm, and determine whether there are any features to be confirmed. If there are no features to be confirmed, the processing feature list is obtained directly. If there are features to be confirmed, the human-machine collaborative confirmation mechanism is executed to obtain the processing feature list. The cost tiered calculation and negotiation module is used to call the preset manufacturing domain knowledge rule base and the preset real-time price database. Based on the structured process instruction list and processing feature list with key manufacturing process parameters, it calculates the technical cost. It obtains non-technical factor information through interaction with the user via a large language model. It judges whether the user's demand is reasonable by combining the manufacturing domain knowledge rule base and the actual execution feasibility. If it is unreasonable, it prompts the user demand and negotiates with him to adjust it until the user demand is reasonable. It then obtains the non-technical user demand and calculates the non-technical cost based on the non-technical user demand. The quotation output module is used to integrate technical and non-technical costs to obtain a total quotation, visualize the total quotation and cost composition, and output a quotation sheet after user confirmation.

[0031] The user interaction layer is used to provide a 3D model visualization window, voice or text input boxes, intelligent dialogue flow interface, quotation result display panel, receive user input, render models and dialogues, and initiate voice recognition operations. The intelligent service layer is built on a microservice architecture and deployed in the cloud or on a local server to provide services such as large language model interaction, geometry processing, cost calculation, and business logic. The data and knowledge layer is used to store structured data, including material parameters, processing time and equipment rates, using a knowledge rule base in the manufacturing field. By accessing dynamic data, including market material prices and outsourcing quotations, a real-time price database is formed. The project and quotation database stores users' historical projects, model files and quotations. The computing support layer provides the hardware and basic software support required for the operation of the user interaction layer, intelligent service layer, and data and knowledge layer. A voice input device is used to convert input voice into text in real time, and the converted text is transmitted to a large language model for interactive services. A 3D model preview device is used to integrate a part visualization window and highlight specific features; A conversational interactive device used to simulate the interaction format of instant messaging software, presenting a dialogue between a user and an artificial intelligence (AI) assistant.

[0032] Example 3 See Figure 3An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a method for intelligent and precise part processing quotation based on a large language model: It acquires unstructured manufacturing requirement information and a 3D part model; uses the large language model to perform semantic parsing on the unstructured manufacturing requirement information and extracts key manufacturing process parameters; combines a pre-set manufacturing domain knowledge rule base to determine whether the unstructured manufacturing requirement information is complete; if not, the large language model generates a clarifying query and interacts with the user in multiple rounds until the unstructured manufacturing requirement information is complete, outputting a structured process instruction list with key manufacturing process parameters; after parsing the 3D part model, it uses a feature recognition algorithm to identify processing features and determines whether there are any features to be confirmed; if no features to be confirmed exist... The process directly yields a list of processing features. If features awaiting confirmation exist, a human-machine collaborative confirmation mechanism is executed to obtain the final processing feature list. A pre-defined manufacturing domain knowledge rule base and a pre-defined real-time price database are invoked. Based on the structured process instruction list with key manufacturing process parameters and the processing feature list, technical costs are calculated. Non-technical factors are obtained through interaction with the user via a large language model. The reasonableness of the user's needs is assessed by combining the manufacturing domain knowledge rule base and actual feasibility. If unreasonable, the user's needs are prompted and adjusted through interactive negotiation until they are reasonable, resulting in non-technical user requirements. Non-technical costs are then calculated based on these requirements. Finally, technical and non-technical costs are integrated to obtain a total quote. The total quote and cost composition are visualized and displayed. After user confirmation, a quote is output.

[0033] Example 4 A computer-readable storage medium stores a computer program that, when executed by a processor, implements a method for intelligent and precise part machining quotation based on a large language model: It acquires unstructured manufacturing requirement information and a 3D part model; uses the large language model to perform semantic parsing on the unstructured manufacturing requirement information and extracts key manufacturing process parameters; combines a pre-set manufacturing domain knowledge rule base to determine whether the unstructured manufacturing requirement information is complete; if not, the large language model generates a clarifying query and interacts with the user in multiple rounds until the unstructured manufacturing requirement information is complete, outputting a structured process instruction list with key manufacturing process parameters; after parsing the 3D part model, it uses a feature recognition algorithm to identify machining features and determines whether there are any features to be confirmed; if no features to be confirmed exist, it directly... The system obtains a list of processing features. If any features require confirmation, a human-machine collaborative confirmation mechanism is executed to obtain the final processing feature list. It then calls upon a pre-defined manufacturing domain knowledge rule base and a pre-defined real-time price database. Based on the structured process instruction list with key manufacturing process parameters and the processing feature list, it calculates the technical costs. Through interaction with the user using a large language model, it obtains information on non-technical factors. Combining this with the manufacturing domain knowledge rule base and actual feasibility, it judges the rationality of the user's needs. If unreasonable, it provides prompts and facilitates interactive negotiation to adjust the needs until they are reasonable, thus obtaining the non-technical user requirements. Based on these requirements, it calculates the non-technical costs. Finally, it integrates the technical and non-technical costs to obtain a total quote. The total quote and cost composition are then visualized and displayed. After user confirmation, a quote is output.

[0034] Example 5 See Figure 4 Where N represents the "no" decision and Y represents the "yes" decision, this method for intelligent and precise parts processing quotation based on a large language model utilizes interactive intelligent manufacturing quotation based on the large language model. Its core lies in constructing a complete closed loop of "intelligent demand analysis - model feature recognition and human-machine collaborative confirmation - precise cost calculation." Through multi-level human-machine interaction, it ensures that the entire process from vague intentions to accurate quotations is reliable, transparent, and efficient. The following sections elaborate on the method, device structure, and integrated system.

[0035] The core of the intelligent pricing method proposed in this embodiment lies in constructing a complete human-machine collaborative closed loop of "intelligent demand analysis - collaborative model feature recognition - accurate cost calculation - visual feedback of results". This method is based on multiple technologies including Large Language Model (LLM), geometric model processing, and intelligent decision-making. (See [link to relevant documentation]). Figure 4 Specifically, it is divided into the following four stages: Step 1: Multimodal Intelligent Analysis and Interactive Clarification of Manufacturing Requirements Based on LLM Step 101, Input Reception: The system supports users to input unstructured natural language processing requirements via voice or text (e.g., "Process 50 aluminum alloy joints, mill a groove in the middle, anodize the surface black, and deliver by next Wednesday"), and supports users to upload 3D part models in STEP (Standard for the Exchange of Product Model Data) format. For example, in this embodiment, the system prompts "Please upload your part model file," and the user inputs a 3D part model in STEP format and specific requirements such as "I want to make 20 parts for the medical field, with a machining accuracy of 0.1mm, the parts are stepped shafts with threaded sections at both ends and a keyway in the middle, used for gear transmission, the material is aluminum-magnesium-silicon alloy 6061, and the surface is treated according to standard," and then begins interactive requirement confirmation.

[0036] Step 102, Semantic Understanding and Key Information Extraction: The LLM module integrated into the system performs deep semantic analysis on the input text to identify and extract key manufacturing process parameters, including material type, processing precision, surface treatment method, special process requirements, delivery time, etc. These key manufacturing process parameters are used to evaluate the processing quotation based on the user's required manufacturing process parameters during the quotation process.

[0037] Step 103, Multi-round Interactive Clarification Mechanism: The LLM module is integrated with a manufacturing domain knowledge rule base to automatically determine the completeness and rationality of user input. If missing information (e.g., material not specified), ambiguity (e.g., "the surface should be brighter"), or potential conflicts are found (e.g., unreasonable roughness markings on medical parts), a clarifying natural language query is proactively generated (e.g., "Please confirm whether the surface treatment requires anodized black?"), prompting the user to confirm or supplement through a front-end dialog box. This process iterates until the requirements are clear.

[0038] Step 104, Output: The LLM module outputs a structured list of process instructions as input to downstream modules.

[0039] Step 2: Feature Recognition and Human-Machine Collaborative Confirmation Step 201, Model Parsing and Feature Recognition: The system receives the STEP 3D part model uploaded by the user, parses it using a geometry processing kernel based on boundary representation (B-Rep, Boundary Representation) (such as PythonOCC (Python Open CASCADE Technology Bindings, a binding library for the open-source 3D geometry kernel Open CASCADE Technology based on the Python language)), and uses an enhanced feature recognition algorithm to identify the machining features (such as holes, slots, cavities, chamfers, etc.).

[0040] Step 202, Uncertainty handling: The user determines whether there are still "unconfirmed" features that need to be modified for each processing feature. If so, proceed to step 203 to execute the human-machine collaborative confirmation mechanism and interact with the LLM to reconfirm the identified processing features. If not, proceed to step 3 to start calculating the cost. Step 203, Human-Machine Collaborative Confirmation Mechanism: For the "Pending Confirmation" feature, the system automatically triggers the interaction process: Step 2031: Highlight the area of ​​the feature in the 3D model on the front-end interface; Step 2032: LLM generates multiple-choice or question-and-answer confirmation prompts (e.g., "This highlighted area may be a 'keyway' or 'relief groove,' please select the correct type?"). For example, when the system performs feature analysis on a 3D model during human-computer interaction for feature recognition, it displays the progress of feature analysis and the time elapsed. When a feature requiring confirmation is found, a prompt is made, indicating the specific feature that needs confirmation. For example, in this embodiment, the system finds a feature requiring confirmation during recognition: a cavity (requiring confirmation): Based on the EWV-AAG algorithm and cavity feature template matching, a concave / through region is detected, but the shape has limited matching degree with a single template; prompting the user to confirm the feature. Here, EW in the EW-AAG algorithm stands for Enhanced Weighted, and AAG stands for Attributed Adjacency Graph. The EW-AAG algorithm is an enhanced weighted attribute adjacency graph commonly used in image processing and model feature recognition. Step 2033: The user confirms by clicking or speaking, and the confirmation result is fed back to the system in real time to correct the feature parameter list.

[0041] Step 2034, Output: Generate an accurate list of processing features and their geometric parameters that have been collaboratively confirmed by the user.

[0042] Step 3: Layered calculation and negotiation of technical and non-technical costs Step 301, Technical Cost Calculation: The cost calculation engine receives the confirmed feature list and process instructions, calls the integrated manufacturing domain knowledge rule base (including cutting parameters, time quotas, material density, etc.) and real-time price database (including material unit price, outsourcing processing quotation, etc.), automatically calculates material costs, processing costs, heat treatment costs, surface treatment costs, etc., and dynamically updates and displays them on the front-end interface.

[0043] Parts processing costs:

[0044] in, This refers to the cost of machining parts, expressed in yuan. The processing time for the parts is expressed in minutes. This is the hourly rate for machine tools, expressed in yuan per hour. Parts and material costs:

[0045] in, This is the cost of parts and materials, in yuan. This refers to the volume of the raw material, in cubic millimeters. The density of the material is expressed in grams per cubic centimeter. The price per kilogram is the unit price of the raw material. Surface treatment cost:

[0046] in, For surface treatment costs, The surface area of ​​the part is expressed in square millimeters. The price listed is the unit price for ordinary surface treatment, expressed in yuan per square meter. Heat treatment cost:

[0047] in, The volume of the part is expressed in cubic millimeters. The unit price for heat treatment is expressed in yuan per kilogram.

[0048] For example, in this embodiment, the technical cost for manufacturing 20 parts for the medical field proposed by the user is calculated, and the detailed total of 2,650 yuan is displayed, listing material cost of 500 yuan, processing cost of 1,200 yuan, surface treatment of 300 yuan, feature processing of 450 yuan, and quality control of 200 yuan.

[0049] Step 302, Non-technical Cost Interaction and Calculation: The LLM module proactively initiates a dialogue, inquiring about non-technical factors such as delivery location, logistics method, and expected lead time. The LLM module has the ability to judge reasonableness. If the user's request is unreasonable (e.g., "ship to a certain location within one day"), for example, regarding transit time requirements, the system uses interfaces with courier service providers such as SF Express and JD.com (based on the courier service requirements, such as standard or SF Express) to determine the time difference between the destination and the origin (provided a friendly prompt and suggested feasible solutions until an agreement is reached). Subsequently, the system calculates logistics and expedited costs based on the negotiated results, including using interfaces with courier service companies such as SF Express for cost estimation. For example, in this embodiment, when a user requests: "I need expedited delivery, preferably tomorrow. The delivery address is a high-tech park. Please use reinforced packaging for better protection. I need a VAT invoice, and there are no other special requirements," the LLM analyzes the user's request and provides a helpful tip: "We must be very frank with you that expecting to receive 20 finished parts tomorrow is impossible from a production perspective. A precision part requires the following essential steps: 1. Programming and machine setup: Write CNC machining programs based on drawings and debug the machine tool. 2. Machining: CNC turning (shaft), milling (keyway), tapping (thread). 3. Post-processing: Surface treatment (such as anodizing, which usually takes several hours). 4. Quality inspection and packaging: Full dimensional inspection and reinforced packaging. This process, even with expedited delivery, normally takes 3-5 working days." The CNC in the tip refers to Computer Numerical Control.

[0050] Step 4: Generating, visualizing, and exporting the quotation. Step 401, Quotation Integration: System integration involves adding up all technical and non-technical costs to generate a total quotation.

[0051] Step 402, Result Visualization: The front-end interface dynamically generates a detailed quotation (including itemized costs, totals, delivery dates, etc.). See [link / reference]. Figure 6 For example, in this embodiment, the quotation specifies the part name, part type, quotation items, and summary, and users can switch to view a pie chart of the technical cost composition. The 3D model view is continuously displayed, and related features can be highlighted.

[0052] Step 403, Interactive Confirmation and Output: The LLM system informs the user that the quotation is complete via a dialog and asks, "Do you want to generate the final quotation?". After the user confirms, the system provides a one-click export function to a PDF (Portable Document Format) quotation, completing the entire intelligent quotation process.

[0053] To implement the above method, this embodiment proposes an intelligent and precise part machining quotation method based on a large language model. It also designs an intelligent and precise part machining quotation system based on a large language model, with cloud-edge-device collaboration. Its hardware and software architecture is as follows: I. The overall system architecture adopts a layered design, see [link / reference] Figure 5 It is divided into four layers in total: First, the user interaction layer (front end): Developed based on Web (World Wide Web) technology, this layer features a responsive interface that runs in the user's browser. It provides a 3D model visualization window, voice / text input boxes, an intelligent dialogue flow interface, and a quotation result display panel. This layer is responsible for receiving user input, rendering models and dialogues, and initiating operations such as voice recognition.

[0054] Second, the intelligent service layer (backend): built on a microservice architecture and deployed in the cloud or on a local server, includes the following core services: LLM Interaction Service: Encapsulates a large language model and is responsible for requirement parsing, dialogue management, and clarification query generation.

[0055] Geometry processing service: Based on geometry kernels such as PythonOCC (Python Open CASCADE Technology Bindings, a Python-based binding library for the open-source 3D geometry kernel Open CASCADE Technology), it is responsible for parsing STEP (Standard for the Exchange of Product Model Data) files and feature recognition.

[0056] Cost calculation service: Integrates knowledge base and database to execute cost calculation logic.

[0057] Business logic service: coordinates the calls to various services and manages business processes and status.

[0058] Third, the data and knowledge layer: includes the following databases: Manufacturing Process Knowledge Base: Stores structured data such as material parameters, processing time, and equipment rates. It is a structured and standardized collection of material parameters and processing data involved in the manufacturing process, serving as a tool for processing quotation.

[0059] Real-time price database: Access dynamic data such as market material prices and outsourcing quotations via API (Application Programming Interface).

[0060] Project and Quotation Database: Stores users' historical projects, model files, quotations, etc.

[0061] Fourth, the computing support layer: provides the hardware and basic software support required for system operation. During deployment, the system can run on a regular workstation (equipped with a modern CPU (Central Processing Unit) and 16GB of memory), and LLM inference can be completed on a consumer-grade GPU (Graphics Processing Unit), demonstrating its friendliness to micro and small enterprises.

[0062] II. Core Interactive Devices and Interfaces: 1) Voice Input Device: The system supports users to input voice through a microphone. The front end integrates the Web Speech API (Web Voice Application Programming Interface) to realize real-time conversion of voice to text. The converted text is automatically filled into the input box and transmitted to the back-end LLM service in real time via the WebSocket protocol (a full-duplex communication protocol based on the Transmission Control Protocol).

[0063] 2) 3D model preview device: The system front end integrates a part visualization window based on PythonOCC, which provides visualization function of the model and supports highlighting specific features according to backend instructions.

[0064] 3) Dialogue Interaction Device: The front-end interface simulates the interaction of instant messaging software. The dialogue between the user and the AI ​​(Artificial Intelligence) assistant is presented in the form of bubbles, accompanied by avatar icons, clearly distinguishing the subjects of the dialogue and providing a natural and smooth interactive experience.

[0065] Based on the above methods and architecture, the intelligent and precise machining quotation system for parts based on a large language model designed in this embodiment has the following workflow: Requirements Input and Parsing: Users log in to the system via a web interface, upload part models, and describe their requirements in natural language. The LLM service parses the data and performs multiple rounds of interaction, outputting structured process instructions.

[0066] Model parsing and collaborative verification: The geometry processing service parses the model and performs feature recognition. The system then confirms the recognition results with the user through the front-end interface, generating the final feature list.

[0067] Cost Calculation and Negotiation: The cost calculation service calculates technical costs based on the feature list and process instructions. The LLM service negotiates non-technical terms with the user and calculates the corresponding fees.

[0068] Quotation generation and delivery: The system generates a detailed quotation and visual charts. After user confirmation, the final quotation (PDF format) is output, completing the service.

[0069] This embodiment of the solution is the first to realize the intelligent, interactive and collaborative manufacturing quotation process, which significantly improves the efficiency and accuracy of quotation, and is especially suitable for small and medium-sized manufacturing enterprises that lack professional process personnel.

[0070] In this embodiment, intelligent parsing and multi-round interaction of manufacturing quotation requirements utilizes LLM as the intelligent hub. It directly understands the user's vague and unstructured processing requirements described in natural language, and proactively judges the completeness and rationality of the requirements through a built-in manufacturing domain knowledge rule base. This results in the generation of clarifying natural language queries and multi-round interaction with the user, ultimately outputting machine-executable, structured process instructions. This method is the first to apply the deep semantic understanding and reasoning capabilities of LLM to the manufacturing quotation field, achieving a fundamental shift from "humans adapting to machines" to "machines understanding humans," significantly lowering the barrier to entry for professional quotation software.

[0071] This embodiment introduces a human-computer collaborative confirmation mechanism for feature recognition results aimed at ensuring the accuracy of quotations. After the system identifies the processing features of the 3D CAD model, if the user believes that the identified features still need modification, the collaborative confirmation mechanism is triggered. The feature area is highlighted on the front-end interface, and easily understandable options or questions are generated using LLM (Local Management Model) to facilitate interaction with the user for the final decision. This method intelligently integrates human judgment into the automated process, ensuring 100% accuracy of the geometric features upon which the quotation is based, and effectively solving the inherent risks of pure algorithmic handling of uncertainty.

[0072] This embodiment presents a cloud-based collaborative manufacturing quotation system for parts based on a large language model. It integrates intelligent interaction, feature processing, and cost calculation, constructing a complete microservice architecture encompassing an intelligent interactive front-end, LLM services, geometric processing services, and cost calculation services. The system seamlessly integrates intelligent parsing of manufacturing quotation requirements with multi-round interactions, utilizing LLM as the intelligent hub. It also integrates a human-machine collaborative confirmation mechanism based on feature recognition results for quotation accuracy with real-time cost calculation, negotiation of non-technical factors, and quotation generation, forming an end-to-end intelligent solution. The system's overall architecture and the collaborative data flow between services (especially the transmission and processing logic of natural language, structured instructions, feature information, and user confirmation signals) ultimately achieve a one-stop intelligent workflow of "requirement input - interactive confirmation - quotation output."

[0073] A low-cost, easily deployable system implementation solution. The core of this solution lies in building the system entirely on open-source technology stacks (such as Dify (Define Modify, an open-source large language model application development platform based on the principles of definition and continuous improvement) and Python OCC), enabling it to run stably without relying on expensive commercial software licenses or high-end hardware resources. The application of this specific combination of open-source technologies in implementing the cloud-based collaborative manufacturing intelligent pricing system is protected, along with its significant economic and social benefits of making it affordable for small and micro enterprises.

[0074] Compared to traditional manual quotation or existing automated quotation software, this embodiment offers the following significant advantages: This embodiment achieves accurate understanding and efficient parsing of unstructured natural language requirements, fundamentally solving the problem of inefficient requirement communication. By introducing a Large Language Model (LLM) as the core of intelligent interaction, the system can directly understand vague or colloquial processing requirements such as "mill a groove, make the surface brighter," and automatically extract key process parameters, completely changing the traditional system's interaction mode that relies on menu selection and structured input. This transforms the previously time-consuming requirement clarification process, which required repeated communication among process engineers, into an automated interaction that takes minutes or even seconds, greatly improving the efficiency and accuracy of requirement communication and effectively solving the semantic gap problem inherent in traditional methods.

[0075] This embodiment ensures the absolute accuracy of the pricing basis by introducing a human-machine collaborative confirmation mechanism: Addressing potential uncertainties in feature recognition, a human-machine interactive confirmation process is proposed. User collaborative decision-making integrates human intelligence into the automated process. This completely avoids the risk of traditional one-way processing systems outputting erroneous results due to algorithmic limitations. The accuracy of the pricing no longer relies entirely on the recognition rate of the backend algorithm, but is built on a reliable foundation of human-machine collaboration, fundamentally solving the problem of inaccurate pricing caused by false detections and missed detections.

[0076] This embodiment provides a highly integrated, low-cost, and easy-to-deploy one-stop solution: This embodiment constructs a complete system integrating "intelligent demand analysis, collaborative model recognition, and accurate cost calculation," implemented using a fully open-source technology stack. The system requires no expensive commercial software licenses and can run on ordinary computing devices, significantly reducing the hardware and software investment threshold and usage costs for enterprises. Simultaneously, the natural language interaction method greatly simplifies operation, enabling small and micro enterprises to enjoy the benefits of cutting-edge AI technology. This effectively resolves the contradiction between the high barriers to entry and high costs of existing professional software and the limited functionality of open-source tools, promoting the democratization of intelligent manufacturing technology.

[0077] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.

[0078] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0079] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0080] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for intelligent and precise machining quotation of parts based on a large language model, characterized in that, include: Obtain unstructured manufacturing requirement information and 3D part models, and use a large language model to perform semantic parsing on the unstructured manufacturing requirement information to extract key manufacturing process parameters. The system uses a pre-defined knowledge rule base in the manufacturing domain to determine whether the unstructured manufacturing requirement information is complete. If it is incomplete, the large language model generates a clarifying query and interacts with the user in multiple rounds until the unstructured manufacturing requirement information is complete. Then, it outputs a list of structured process instructions with key manufacturing process parameters. After analyzing the three-dimensional part model, a feature recognition algorithm is used to identify the processing features and determine whether there are any features to be confirmed. If there are no features to be confirmed, the processing feature list is obtained directly. If there are features to be confirmed, a human-machine collaborative confirmation mechanism is executed to obtain the processing feature list. The system invokes the preset manufacturing domain knowledge rule base and the preset real-time price database, and calculates the technical cost based on the structured process instruction list with key manufacturing process parameters and the processing feature list. It then interacts with the user through the large language model to obtain non-technical factor information, and combines this with the manufacturing domain knowledge rule base and actual execution feasibility to determine whether the user's demand is reasonable. If unreasonable, it provides a prompt and engages in interactive negotiation to adjust the demand until it becomes reasonable, thus obtaining the non-technical user demand. Based on this non-technical user demand, the non-technical cost is calculated. The total price is obtained by integrating the technical and non-technical costs. The total price and cost composition are then visualized and displayed. After user confirmation, a price list is generated.

2. The intelligent and precise machining quotation method for parts based on a large language model according to claim 1, characterized in that, The input formats for the unstructured manufacturing requirements information include voice input and text input; The format of the three-dimensional part model is the STEP format, the product model data exchange standard. The key manufacturing process parameters include material type, machining accuracy, surface treatment method, special process requirements, and delivery time; The incompleteness of unstructured manufacturing demand information includes situations where information is missing, ambiguous, or conflicting.

3. The intelligent and precise machining quotation method for parts based on a large language model according to claim 1, characterized in that, When parsing the three-dimensional part model, a geometric processing kernel based on boundary representation B-Rep is used. The feature recognition algorithm adopts an enhanced feature recognition algorithm; The machining features include holes, grooves, cavities, and chamfers.

4. The intelligent and precise machining quotation method for parts based on a large language model according to claim 1, characterized in that, If there are features to be confirmed, a human-machine collaborative confirmation mechanism is executed, including: Mark the areas of features to be confirmed in the visualization interface of the three-dimensional part model to obtain the areas to be confirmed; For the area to be confirmed, the large language model generates multiple-choice or question-and-answer confirmation prompts to interact with the user and obtain the confirmation result. Based on the confirmation results, the processing feature parameters are corrected to obtain a collaboratively confirmed list of processing features; Output a list of collaboratively confirmed processing features.

5. The intelligent and precise machining quotation method for parts based on a large language model according to claim 1, characterized in that, The manufacturing domain knowledge rule base includes cutting parameters, time quotas, and material density; The preset real-time price database includes material unit prices and outsourcing processing quotations; The technical costs include material costs, processing costs, heat treatment costs, and surface treatment costs; The non-technical factors include delivery location, logistics method, and expected delivery time; The non-technical costs include logistics fees and expedited processing fees.

6. The intelligent and precise machining quotation method for parts based on a large language model according to claim 1, characterized in that, The visualization includes a cost breakdown chart, and the quotation includes itemized costs, total costs, and delivery dates. The quotation is output in PDF format.

7. A parts intelligent precision machining quotation system based on a large language model, characterized in that, include: The intelligent demand analysis and interactive clarification module is used to acquire unstructured manufacturing demand information and 3D part models. It uses a large language model to perform semantic analysis on the unstructured manufacturing demand information and extract key manufacturing process parameters. It combines a preset manufacturing domain knowledge rule base to determine whether the unstructured manufacturing demand information is complete. If it is incomplete, the large language model generates clarification queries and interacts with the user in multiple rounds until the unstructured manufacturing demand information is complete. Finally, it outputs a list of structured process instructions with key manufacturing process parameters. The feature recognition and human-machine collaborative confirmation module is used to analyze the three-dimensional part model, identify the processing features using a feature recognition algorithm, and determine whether there are any features to be confirmed. If there are no features to be confirmed, the processing feature list is obtained directly. If there are features to be confirmed, the human-machine collaborative confirmation mechanism is executed to obtain the processing feature list. The cost tiered calculation and negotiation module is used to call the preset manufacturing domain knowledge rule base and the preset real-time price database, and calculate the technical cost based on the structured process instruction list with key manufacturing process parameters and the processing feature list; it also uses the large language model to interact with the user to obtain non-technical factor information, and combines the manufacturing domain knowledge rule base and actual execution feasibility to judge whether the user's demand is reasonable. If it is unreasonable, it prompts the unreasonable user demand and negotiates and adjusts it interactively until the user demand is reasonable, thus obtaining the non-technical user demand, and calculating the non-technical cost based on the non-technical user demand. The quotation output module is used to integrate the technical and non-technical costs to obtain a total quotation, visualize the total quotation and cost composition, and output a quotation sheet after user confirmation.

8. The intelligent and precise machining quotation system for parts based on a large language model according to claim 7, characterized in that, Also includes: The user interaction layer is used to provide a 3D model visualization window, voice or text input boxes, intelligent dialogue flow interface, quotation result display panel, receive user input, render models and dialogues, and initiate voice recognition operations. The intelligent service layer is built on a microservice architecture and deployed in the cloud or on a local server to provide services such as large language model interaction, geometry processing, cost calculation, and business logic. The data and knowledge layer is used to store structured data, including material parameters, processing time and equipment rates, using a knowledge rule base in the manufacturing field. By accessing dynamic data, including market material prices and outsourcing quotations, a real-time price database is formed. The project and quotation database stores users' historical projects, model files and quotations. The computing support layer provides the hardware and basic software support required for the operation of the user interaction layer, intelligent service layer, and data and knowledge layer. A voice input device is used to convert input voice into text in real time, and the converted text is transmitted to the large language model for interactive services. A 3D model preview device is used to integrate a part visualization window and highlight specific features; A conversational interactive device used to simulate the interaction format of instant messaging software, presenting a dialogue between a user and an artificial intelligence (AI) assistant.

9. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the intelligent and precise machining quotation method for parts based on a large language model as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the intelligent and precise part machining quotation method based on a large language model as described in any one of claims 1-6.

Images