A method and system for formulating a numerical control machine tool number-real fusion test scheme based on a large language model

Abstract

This application discloses a method and system for formulating a numerical-physical fusion test scheme for CNC machine tools based on a large language model. The method includes: using a constructed test requirement analysis agent to generate a set of topic prompt words based on the input test question and multiple preset prompt dimensions; and using a large language model to generate digital test schemes, physical test schemes, and numerical-physical fusion schemes based on the topic prompt word set and a pre-constructed test knowledge base. The test requirement analysis agent is constructed using the ReAct framework based on a large language model; the test knowledge base is constructed based on multiple preset knowledge dimensions, consisting of multiple sets of knowledge units and the relationships between these sets, with one knowledge dimension corresponding to one set of knowledge units. This application enables the intelligent formulation of numerical-physical fusion test schemes for CNC machine tools even when users have unclear understanding of numerical-physical fusion testing or vague test requirement descriptions.

Description

Technical Field

[0001] This application relates to the field of CNC machine tool testing, specifically to a method and system for developing a CNC machine tool numerical-real fusion testing scheme based on a large language model. Background Technology

[0002] The statements in this section are merely to provide background information in relation to this application to aid in understanding it, and such background information does not necessarily constitute prior art.

[0003] In the field of high-end CNC machine tool testing, the test plan is the key carrier connecting test requirements, test resources, and the test implementation process. Its rationality and completeness directly affect the accuracy of test results, the efficiency of the test process, and the credibility of test conclusions. With the development of technologies such as digital twins, mechanism modeling, physical experiments, and data fusion, CNC machine tool testing has gradually evolved from a single physical experiment mode to a collaborative mode of digital testing, physical testing, and data-physical fusion testing. Among these, there are complex multi-dimensional coupling relationships between test objects, performance indicators, typical working conditions, evaluation indicators, digital models, physical testing methods, and data-physical fusion tasks. Different test requirements necessitate matching different digital test tasks, physical test tasks, and data-physical fusion strategies. Therefore, how to quickly generate structured, standardized, and executable data-physical fusion test plans for practical problems has become a critical issue that urgently needs to be addressed in the field of intelligent CNC machine tool testing. Summary of the Invention

[0004] To address the technical challenges of understanding test requirements, organizing test knowledge, and intelligently generating test schemes during the generation of CNC machine tool fusion test plans, this application constructs an intelligent agent for analyzing CNC machine tool fusion test requirements based on a large language model, establishes a multi-dimensional test knowledge base, and achieves intelligent generation of CNC machine tool fusion test schemes. To achieve the above objectives, this application adopts the following technical solution:

[0005] The objective of this application is achieved through the following technical solution:

[0006] According to the first aspect of this application, a method for formulating a numerical-physical fusion test scheme for CNC machine tools based on a large language model is provided, comprising: using a constructed test requirement analysis agent to generate a set of topic prompt words based on the input test question and multiple preset prompt dimensions; and using the large language model to generate a digital test scheme, a physical test scheme, and a numerical-physical fusion scheme based on the set of topic prompt words and a pre-constructed test knowledge base, to obtain a complete set of test schemes; wherein, the test requirement analysis agent is constructed based on a large language model using the ReAct framework; and the test knowledge base is constructed based on multiple preset knowledge dimensions, which consists of multiple sets of knowledge units and the relationships between the sets of knowledge units, with one knowledge dimension corresponding to one set of knowledge units.

[0007] Preferably, the prompt dimensions include at least: test object, performance indicators, test conditions, evaluation indicators, digital test tasks, physical test tasks, and data-real fusion tasks.

[0008] Preferably, the requirement analysis agent includes a problem understanding agent, a requirement inference agent, and a data-real task recognition agent; wherein, the problem understanding agent is configured to perform a problem understanding process based on the test problem, including problem phenomenon identification, test component localization, and test performance inference, to obtain a problem understanding result; the requirement inference agent is configured to perform a problem requirement transformation process based on the problem understanding result to obtain a test requirement result, and output test object prompts, performance indicator prompts, test condition prompts, and evaluation indicator prompts based on the test requirement result and preset criteria for prompts; the data-real task recognition agent is configured to perform a task recognition process based on the test problem and the test requirement result, and output digital test task prompts, physical test task prompts, and data-real fusion task prompts; wherein, the test object prompts, performance indicator prompts, test condition prompts, evaluation indicator prompts, digital test task prompts, physical test task prompts, and data-real fusion task prompts constitute a set of theme prompts.

[0009] Preferably, the problem understanding process includes problem phenomenon identification, test component location, and test performance inference; the problem requirement transformation process includes problem-to-test requirement mapping, requirement standardization, and requirement structure generation; the task identification process includes identifying digital and physical tasks, analyzing the mapping relationship between digital and physical tasks to identify digital-physical fusion tasks, and if a mismatch is found between digital and physical tasks during the identification process of digital-physical fusion tasks, feedback correction of digital and physical tasks is provided.

[0010] Preferably, the test object prompts are used to characterize the system under test and its key components; the performance index prompts are used to characterize the performance type to be tested; the test condition prompts are used to characterize the cutting and machining conditions during test execution; the evaluation index prompts are used to characterize the measurement standards for test results; the digital test task prompts are used to characterize the construction and simulation tasks of the digital model; the physical test task prompts are used to characterize the methods, equipment, and data acquisition, processing, and analysis tasks of physical testing; and the data-real fusion task prompts are used to characterize the fusion task of physical and digital test objects, scenarios, processes, and results.

[0011] Preferably, the knowledge dimensions include at least: test objects, performance indicators, test conditions, digital tests, physical tests, and data-real fusion.

[0012] Preferably, based on the aforementioned keyword set and the constructed test knowledge base, a large language model is used to generate digital test schemes, physical test schemes, and data-real fusion schemes, respectively. This includes: querying the test knowledge base for knowledge entries containing prompts for test objects, performance indicators, test conditions, evaluation indicators, digital test tasks, physical test tasks, and data-real fusion tasks, to obtain corresponding digital test knowledge, physical test knowledge, and data-real fusion test knowledge; using the digital test knowledge as prompts, combined with the prompts for test objects, performance indicators, test conditions, and evaluation indicators, to generate a digital test scheme using a large language model; using the physical test knowledge as prompts, combined with the prompts for test objects, performance indicators, test conditions, and evaluation indicators, to generate a physical test scheme using a large language model; and using the data-real fusion test knowledge as prompts, combined with the prompts for test objects, performance indicators, test conditions, and evaluation indicators, to generate a data-real fusion test scheme using a large language model.

[0013] Preferably, the test knowledge base is constructed as follows: a corresponding prompt word template is determined for each knowledge dimension; multiple knowledge entries corresponding to the knowledge dimension are generated using a large language model based on the prompt word template, and the multiple knowledge entries constitute a knowledge unit set corresponding to the knowledge dimension; the association relationship between knowledge unit sets is determined based on the association relationship between knowledge entries in different knowledge unit sets, wherein the association relationship includes at least one of the following: applicability relationship, dependency relationship, verification relationship, mapping relationship, and compensation relationship; and the test knowledge base is constructed based on all knowledge unit sets and the association relationship between all knowledge unit sets.

[0014] According to a second aspect of this application, a system for formulating numerical-physical fusion test schemes for CNC machine tools based on a large language model is provided, including a topic prompt word generation module and a test scheme generation module. The topic prompt word generation module is configured to use a constructed test requirement analysis agent to generate a set of topic prompt words based on the input test question and multiple preset prompt dimensions. The test requirement analysis agent is constructed using the ReAct framework based on a large language model. The test scheme generation module is configured to generate digital test schemes, physical test schemes, and numerical-physical fusion schemes respectively using the large language model based on the topic word set and a pre-constructed test knowledge base, to obtain a complete set of test schemes. The test knowledge base is constructed based on multiple preset knowledge dimensions, consisting of multiple knowledge unit sets and the relationships between these knowledge unit sets, with one knowledge dimension corresponding to one knowledge unit set.

[0015] According to a third aspect of this application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of the first aspect of this application.

[0016] According to a fourth aspect of this application, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method of the first aspect of this application.

[0017] Compared with the prior art, the main advantages of this application are:

[0018] This application constructs a multi-dimensional test knowledge base, achieving a systematic organization, standardized definition, and multi-dimensional relational expression of professional knowledge in the field of CNC machine tool digital-physical fusion testing. This fundamentally solves the pain points of fragmented knowledge, non-standardized retrieval, and incomplete coverage in traditional test solution development. The knowledge unit set integrates scattered professional content such as test standards, equipment parameters, model methods, and operating condition data into a knowledge system that can be efficiently invoked by a large language model through unified structured fields and semantic association networks. This significantly improves the standardization, completeness, and traceability of test knowledge retrieval, providing a solid and reliable knowledge foundation for test solution generation. Through the intelligent parsing and structured transformation of test requirements by the requirements analysis agent, combined with precise matching from the multi-dimensional knowledge base, it can automatically generate complete test solutions covering the entire chain of digital simulation, physical experimentation, and digital-physical fusion testing, even in scenarios where users have unclear test understanding or vague requirements. This significantly improves the efficiency, structure, and intelligence level of CNC machine tool digital-physical fusion test solution generation. Driven by a large language model, this system enables the collaborative generation of digital testing schemes, physical testing schemes, and data-physical fusion schemes throughout the entire process. This completely transforms the traditional, experience-dependent, inefficient, and poorly structured approach of manually developing schemes. Furthermore, the targeted generation of the data-physical fusion scheme achieves deep coupling between the digital model and physical testing, effectively ensuring the consistency of test results between digital and physical models and their engineering applicability. This provides an intelligent and standardized solution for CNC machine tool performance testing, condition assessment, and optimization iteration, possessing extremely high engineering application value and promising prospects for widespread adoption. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0020] Figure 1 A schematic diagram illustrating the conceptual principle of a numerical-real fusion testing scheme for CNC machine tools based on a large language model, according to an embodiment of this application;

[0021] Figure 2 This is a flowchart illustrating a method for developing a numerical-to-real-data fusion testing scheme for CNC machine tools based on a large language model, according to an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. Furthermore, the technical features involved in the various embodiments described below can be combined with each other as long as they do not conflict with each other.

[0023] During their research on CNC machine tool testing, the inventors discovered that existing technologies still face many challenges in the process of intelligently generating practical data-real fusion testing solutions. Firstly, regarding the expression of test requirements, existing requirements are mostly derived from fault phenomenon descriptions, empirical requirement descriptions, or unstructured text records, resulting in inconsistent expressions, ambiguous semantics, and vague requirement boundaries. This makes it difficult to achieve a standardized mapping from test problems to test objects, performance indicators, typical operating conditions, and test tasks. Secondly, in terms of test knowledge organization, existing test-related knowledge is usually scattered across test standards, test manuals, expert experience, model documents, and historical test cases. The multi-dimensional knowledge of the tested object, performance indicators, typical operating conditions, digital models, physical testing, and data-real fusion lacks unified organization and related expression, making it difficult to support the accurate understanding and application of complex test scenarios by large models. Third, in terms of test plan generation, the existing test plan development process mainly relies on human experience for compilation and adjustment. There is a lack of a unified collaborative generation mechanism among digital test plans, physical test plans, and data-physical fusion plans, resulting in low efficiency in plan generation, insufficient structuring, and difficulty in quickly reusing and optimizing plans for different test scenarios.

[0024] To better understand this application, we will first introduce the overall concept of the formulation method provided by one embodiment of this application. Figure 1 This is a schematic diagram illustrating the conceptual principle of a data-real fusion testing scheme based on a large language model, according to an embodiment of this application. Figure 1As shown, the overall process is divided into three parts: constructing an intelligent agent for data-real fusion test requirements analysis, constructing a data-real fusion test knowledge base, and intelligently generating data-real fusion test solutions. The construction of the intelligent agent for test requirements analysis includes building a problem understanding agent, a requirement inference agent, and a data-real task recognition agent. These three agents collaborate to complete the initial analysis of test requirements. Based on the analysis results of the multiple agents, a set of topic prompt words is generated to create a data-real fusion test solution for CNC machine tools. Based on the set of topic prompt words, a metadata file for the structured requirements of data-real fusion test for CNC machine tools is established, transforming unstructured test requirements into standardized, structured metadata and unifying the requirement input format. The construction of the data-real fusion test knowledge base includes: defining data-real fusion test knowledge in units, constructing a set of knowledge units containing six categories: the test object under test, performance indicators, typical working conditions, digital tests, physical tests, and data-real fusion, and clarifying the structured fields of each type of knowledge unit set. Based on the defined knowledge unit sets, a large language model is called to generate structured knowledge entries corresponding to each type of knowledge unit set, completing the content filling of the knowledge base. Based on knowledge entries, relationships are established between different sets of knowledge units, ultimately forming a multi-dimensional knowledge base containing six types of knowledge unit sets and their relationships. The intelligent generation of data-physical fusion test schemes includes: based on the set of topic prompts and the multi-dimensional knowledge base, calling a large language model to generate digital test schemes, physical test schemes, and data-physical fusion schemes for CNC machine tool data-physical fusion testing. These digital test schemes, physical test schemes, and data-physical fusion schemes are then summarized to form a final, complete set of CNC machine tool data-physical fusion test schemes, completing the entire closed-loop process.

[0025] The following section will provide a detailed description of the method for developing a data-real fusion testing scheme based on a large language model, according to embodiments of this application.

[0026] Figure 2 This is a flowchart illustrating a method for developing a data-real fusion testing scheme based on a large language model according to an embodiment of this application. Figure 2 As shown, the method includes:

[0027] Step S201: Utilize the constructed test requirement analysis agent to generate a set of topic prompts based on the input test question and multiple preset prompt dimensions. The test requirement analysis agent is constructed using a Reasoning + Acting (ReAct) framework based on a large language model. In some embodiments, the prompt dimensions include at least: test object, performance metrics, test conditions, evaluation metrics, digital test tasks, physical test tasks, and data-real fusion tasks.

[0028] Step S202: Based on the keyword set and the pre-built test knowledge base, generate digital test schemes, physical test schemes, and data-real fusion schemes using a large language model to obtain a complete set of test schemes. The test knowledge base is constructed based on multiple preset knowledge dimensions, consisting of multiple sets of knowledge units and the relationships between these sets; one knowledge dimension corresponds to one set of knowledge units.

[0029] This application's embodiments, by constructing a multi-dimensional test knowledge base, systematically organize, standardize, and express multi-dimensional related professional knowledge in the field of CNC machine tool digital-physical fusion testing. This fundamentally solves the pain points of fragmented knowledge, non-standardized retrieval, and incomplete coverage in traditional test solution development. The knowledge unit set, through unified structured fields and semantic association networks, integrates scattered professional content such as test standards, equipment parameters, model methods, and operating condition data into a knowledge system that can be efficiently invoked by a large language model. This significantly improves the standardization, completeness, and traceability of test knowledge retrieval, providing a solid and reliable knowledge foundation for test solution generation. Through the intelligent parsing and structured transformation of test requirements by the requirements analysis agent, combined with precise matching from the multi-dimensional knowledge base, it can automatically generate complete test solutions covering the entire chain of digital simulation, physical experimentation, and digital-physical fusion testing, even in scenarios where users have unclear test understanding or vague requirements descriptions. This significantly improves the efficiency, structure, and intelligence level of CNC machine tool digital-physical fusion test solution generation. Driven by a large language model, the system enables the collaborative generation of digital testing schemes, physical testing schemes, and data-real fusion schemes throughout the entire process, completely changing the traditional situation where manually formulated schemes rely on experience, are inefficient, and lack structure.

[0030] In some embodiments, step S201 mainly includes: constructing a test requirements analysis agent based on a large language model using the ReAct agent framework. .in, This is a problem-understanding agent for CNC machine tool fusion testing, covering the problem understanding process of "problem phenomenon identification → test component location → test performance inference". The intelligent agent for inferring the testing requirements of CNC machine tools through the fusion of numerical and real data covers the process of transforming problems into requirements, namely, "mapping problems into testing requirements → standardizing requirements → generating structured requirements". It is an intelligent agent for identifying digital and physical tasks, covering a closed loop of task organization: "digital test task identification → physical test task identification → digital-physical fusion task identification". That is, it identifies digital and physical tasks, analyzes the mapping relationship between digital and physical tasks to identify digital-physical fusion tasks, and provides feedback to correct digital and physical tasks if a mismatch is found between digital and physical tasks during the identification process of digital-physical fusion tasks.

[0031] In some embodiments, the problem understanding agent is configured to perform a problem understanding process based on the test problem, including problem phenomenon identification, test component localization, and test performance inference, to obtain a problem understanding result. The requirement inference agent is configured to perform a problem requirement transformation process based on the problem understanding result to obtain test requirement results, and output test object prompts, performance indicator prompts, test condition prompts, and evaluation indicator prompts based on the test requirement results and preset criteria for prompts. The data-real task recognition agent is configured to perform a task recognition process based on the test problem and test requirement results, and output digital test task prompts, physical test task prompts, and data-real fusion task prompts.

[0032] In some embodiments, the preset basis for prompt words is mainly set according to the application industry and field. For example, in the field of CNC machine tool testing, the preset basis for test object prompt words is the classification of typical CNC machine tool components, and the preset basis for performance index prompt words is the core performance type of CNC machine tool. The preset basis for test condition prompt words is typical machining condition mode. The preset basis for evaluation index prompt words is industry-standard test index. The preset basis for digital test task prompt words is digital modeling technology type. The preset basis for physical test task prompt words is physical test element classification. The preset basis for digital-real fusion task prompt words is key technology of virtual-real collaboration.

[0033] In some embodiments, test object prompts are used to characterize the system under test and its key components; performance index prompts are used to characterize the type of performance to be tested; test condition prompts are used to characterize the cutting and machining conditions during test execution; evaluation index prompts are used to characterize the measurement standards for test results; digital test task prompts are used to characterize the construction and simulation tasks of digital models; physical test task prompts are used to characterize the methods, equipment, and data acquisition, processing, and analysis tasks of physical testing; and data-real fusion task prompts are used to characterize the fusion task of physical and digital test objects, scenarios, processes, and results.

[0034] In the field of CNC machine tool fusion testing, the test requirement analysis intelligent agent generates a set of topic prompt words. ,in, The test object prompts include spindle system, basic structural components, feed system, rotary table system, etc. These are performance indicator terms, including dynamic performance, service performance, and thermal stability. The test condition prompts include high-speed cutting, continuous operation, no-load operation, and load operation. The evaluation indicators include terms such as temperature rise, thermal error, vibration acceleration, surface roughness, and repeatability error. This refers to prompts for digital testing tasks, including mechanistic models, data models, digital twin models, and digital testing tasks. These refer to the prompts for physical testing tasks, including test methods, test equipment, test standards, and data acquisition tasks. The prompts for the index-real fusion task include data-real mapping relationships and data mechanism fusion.

[0035] To transform unstructured test requirements into standardized, structured metadata, a CNC machine tool CNC-real fusion test requirement structure metadata file, `requirements.json`, can be created. In some embodiments, this is based on a set of topic prompt words. Constructing a set of data-real integration test requirements fields in R FS .

[0036] R FS {'input_type', 'problem_phenomenon', 'test_object', 'component', 'performance_goal', 'operating_condition', 'evaluation_indicator', 'digital_task', 'physical_task', 'fusion_task', 'missing_information'}.

[0037] Among them, input_type represents the input type field, problem_phenomenon represents the problem phenomenon field, test_object represents the test object field, component represents the key component field, performance_goal represents the performance index field, operating_condition represents the test condition field, evaluation_indicator represents the evaluation index field, digital_task represents the digital test task field, physical_task represents the physical test task field, fusion_task represents the digital-real fusion task field, and missing_information represents the missing information field.

[0038] In some embodiments, the test knowledge base in step S202 can be constructed as follows: A corresponding prompt word template is determined for each knowledge dimension. Multiple knowledge entries corresponding to the knowledge dimension are generated using a large language model based on the prompt word templates, and these multiple knowledge entries constitute a knowledge unit set corresponding to the knowledge dimension. The relationships between knowledge unit sets are determined based on the relationships between knowledge entries in different knowledge unit sets, wherein the relationships include at least one of the following: applicability, dependency, verification, mapping, and compensation. The test knowledge base is constructed based on all knowledge unit sets and the relationships between all knowledge unit sets. In some embodiments, the knowledge dimension includes at least: test object, performance index, test condition, digital model, physical test, and data-real fusion.

[0039] In some embodiments, the process of building a test knowledge base specifically includes:

[0040] 1. Definition of the set of knowledge units for the integration of numerical and real-world data testing:

[0041] Data-Real Integration Test Knowledge Base , For the knowledge unit set of the test object, It is a collection of knowledge units for performance indicators. This is a collection of knowledge units for testing operating conditions. A collection of digital test knowledge units. It is a collection of knowledge units for physics testing. It is a collection of knowledge units that integrate numbers and reality.

[0042] Each type of knowledge unit set It consists of several knowledge items, namely .

[0043] Among them, the One knowledge entry . The name of the knowledge entry. For the categories of knowledge entries, These are key attributes or parameters of a knowledge entry. For potential related objects or indicators of knowledge entries, The units and range of values ​​for each knowledge item.

[0044] 2. Generation of knowledge items for data-real integration testing:

[0045] According to the formula Calling the large language model The API is a collection of knowledge units for each type. Generate structured knowledge entries. Among them, The first generated for the large model A structured knowledge item, For knowledge unit sets The designed prompt template includes fields, attributes, and formatting. Structured knowledge entries. fields and The defined fields are consistent, that is .

[0046] 3. Establishing knowledge links for data-real integration testing:

[0047] Relationships between knowledge unit sets ;

[0048] in, Represents a set of knowledge units The Middle One entry, Represents a set of knowledge units The Middle One entry, express and The semantic or logical relationships between them, including "applicability," "dependency," "verification," "mapping," "compensation," etc. for The total number of knowledge items in the Chinese language. for The total number of knowledge entries in the text.

[0049] 4. Establish a multi-dimensional test knowledge base :

[0050] It contains six types of knowledge unit sets and their relationships.

[0051] In some embodiments, the digital test plan in step S202 can be generated from a test knowledge base. The query contains prompts for the test object. Performance indicator prompts Test status prompts Evaluation indicator prompts and digital test task prompts The knowledge items will yield the corresponding numerical test knowledge. This involves generating numerical test schemes using large language models. Specifically, it includes:

[0052] According to the formula Calling the large language model API generates digital test solutions. The first generation of large language models A numerical test plan.

[0053] Among them, the Numerical test scheme :

[0054] ;

[0055] ;

[0056] in, The number of digital test cases generated. This is the name of the digital test scheme. For digital test task type, The name of the object being tested. For the set of applicable performance indicator names, To accommodate the set of typical operating condition names, For input parameters or input data, This is for outputting results or data.

[0057] Digital Test Solution Collection ;

[0058] In some embodiments, the physical test plan in step S202 can be generated from a test knowledge base in the following manner: The query contains prompts for the test object. Performance indicator prompts Test status prompts Evaluation indicator prompts And physics test task prompts The knowledge entries will lead to the corresponding physics test knowledge. The system utilizes large language models to generate physical testing schemes. Specifically, this includes:

[0059] According to the formula Calling the large language model The API generates physical test schemes. The first generation of large language models A physical testing plan.

[0060] Among them, the Physical test plan :

[0061] ;

[0062] ;

[0063] in, The number of physical test schemes generated. This is the name of the physical testing scheme. This is a physical test task type. The name of the object being tested. For the set of applicable performance indicator names, To accommodate the set of typical operating condition names, For testing methods, For testing equipment, The measurement point layout and sampling parameters are specified.

[0064] Collection of physical testing solutions .

[0065] In some embodiments, the data-real fusion test scheme in step S202 can be generated from a test knowledge base in the following manner: Test object knowledge unit set Test Condition Knowledge Unit Collection and a collection of performance indicator knowledge units The query contains prompts for the test object. Performance indicator prompts Test status prompts Evaluation indicator prompts and prompts for data-real fusion tasks The knowledge entries yield the corresponding data-real fusion test knowledge. This involves generating a data-real fusion test scheme using a large language model. Specifically, it includes:

[0066] According to the formula Calling the large language model API-generated data-real fusion solution The first generation of large language models A solution for integrating data and quantity.

[0067] Among them, the Data integration solution :

[0068] ;

[0069] ;

[0070] in, The number of generated data-real fusion schemes, The name of the data-real integration solution. The name of the object being tested. To associate the set of numerical test scheme names, To associate the set of physical test scheme names, This relates to the alignment and mapping between digital test results and physical test results. For the real deviation compensation and fusion strategy, Output data for the fusion of data and reality.

[0071] Collection of Data-Real Integration Solutions .

[0072] In some embodiments, step S202 may further include: establishing a set of numerical-real fusion test schemes for CNC machine tools. .in, A collection of digital test solutions. A collection of physical testing schemes. A collection of data-real integration solutions. This is the final set of CNC machine tool numerical and real-data fusion test schemes.

[0073] Based on the same inventive concept, one embodiment of this application provides a system for formulating a numerical-physical fusion test scheme for CNC machine tools based on a large language model, including a topic prompt word generation module and a test scheme generation module. The topic prompt word generation module is configured to use a constructed test requirement analysis agent to generate a set of topic prompt words based on the input test question and multiple preset prompt dimensions. The test requirement analysis agent is constructed using the ReAct framework based on a large language model. The test scheme generation module is configured to generate digital test schemes, physical test schemes, and numerical-physical fusion schemes respectively using the large language model based on the topic word set and a pre-constructed test knowledge base, to obtain a complete set of test schemes. The test knowledge base is constructed based on multiple preset knowledge dimensions, consisting of multiple knowledge unit sets and the relationships between these knowledge unit sets, with one knowledge dimension corresponding to one knowledge unit set. Since the system embodiment is basically similar to the method embodiment, the details of the relevant technical features and the effects of implementation can be found in the corresponding descriptions of the method embodiment provided above.

[0074] This application's embodiments, by constructing a multi-dimensional test knowledge base, systematically organize, standardize, and express multi-dimensional related professional knowledge in the field of CNC machine tool digital-physical fusion testing. This fundamentally solves the pain points of fragmented knowledge, non-standardized retrieval, and incomplete coverage in traditional test solution development. The knowledge unit set, through unified structured fields and semantic association networks, integrates scattered professional content such as test standards, equipment parameters, model methods, and operating condition data into a knowledge system that can be efficiently invoked by a large language model. This significantly improves the standardization, completeness, and traceability of test knowledge retrieval, providing a solid and reliable knowledge foundation for test solution generation. Through the intelligent parsing and structured transformation of test requirements by the requirements analysis agent, combined with precise matching from the multi-dimensional knowledge base, it can automatically generate complete test solutions covering the entire chain of digital simulation, physical experimentation, and digital-physical fusion testing, even in scenarios where users have unclear test understanding or vague requirements descriptions. This significantly improves the efficiency, structure, and intelligence level of CNC machine tool digital-physical fusion test solution generation. Driven by a large language model, the system enables the collaborative generation of digital testing schemes, physical testing schemes, and data-real fusion schemes throughout the entire process, completely changing the traditional situation where manually formulated schemes rely on experience, are inefficient, and lack structure.

[0075] The embodiments provided in this application can also be computer program products. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this application.

[0076] Computer-readable storage media can be tangible devices that hold and store instructions for use by an instruction execution device. Computer-readable storage media can be, for example, including but not limited to, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof.

[0077] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system and modules described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0078] In the embodiments provided in this application, it should be understood that the division of modules (or units) is merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the coupling or direct coupling or communication connection between units may be through some interfaces, or indirect coupling or communication connection between units, or it may be an electrical, mechanical or other form of connection.

[0079] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The aforementioned units can be implemented in hardware or as software functions.

[0080] The various embodiments in this application are described in a progressive manner, with each embodiment focusing on the differences from other embodiments or implementation methods. Similar or identical parts between the various embodiments of this application can be referred to mutually. The implementation principles and technical effects of the inventive concept can be mutually referenced, and will not be repeated here. Where there is no conflict, the various embodiments or implementation methods in this application can be combined with each other.

[0081] It should be noted that although the steps are described in a specific order above, it does not mean that the steps must be executed in the above specific order. In fact, some of these steps can be executed concurrently or even in a different order, as long as the required function can be achieved.

[0082] This application uses specific embodiments to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the solution and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for formulating a numerical control machine tool number-real fusion test scheme based on a large language model, characterized in that, include: Using the constructed test requirements analysis agent, a set of topic prompt words is generated based on the input test questions and multiple preset prompt dimensions; Based on the aforementioned keyword set and the pre-built test knowledge base, a large language model is used to generate digital test schemes, physical test schemes, and data-real fusion schemes to obtain a complete set of test schemes. The test requirement analysis agent is built using the ReAct framework based on a large language model; the test knowledge base is built according to multiple preset knowledge dimensions, which consists of multiple sets of knowledge units and the relationships between the sets of knowledge units, with one knowledge dimension corresponding to one set of knowledge units.

2. The method of claim 1, wherein, The prompt dimensions include at least: test object, performance indicators, test conditions, evaluation indicators, digital test tasks, physical test tasks, and data-real fusion tasks.

3. The method of claim 2, wherein, The test requirement analysis agent includes a problem understanding agent, a requirement inference agent, and a data-real-task recognition agent; among which... The problem understanding agent is configured to perform a problem understanding process based on the test problem, including problem phenomenon identification, test component localization, and test performance inference, to obtain the problem understanding result; The requirement inference agent is configured to perform a problem requirement transformation process based on the problem understanding result to obtain the test requirement result, and output test object prompt words, performance indicator prompt words, test condition prompt words and evaluation indicator prompt words based on the test requirement result and the preset basis of prompt words; The digital-real task recognition agent is configured to perform a task recognition process based on the test question and the test requirement results, and output digital test task prompt words, physical test task prompt words and digital-real fusion task prompt words; The set of thematic prompts consists of prompts for test objects, performance indicators, test conditions, evaluation indicators, digital test tasks, physical test tasks, and data-real fusion tasks.

4. The method of claim 3, wherein, The problem understanding process includes problem phenomenon identification, test component location, and test performance inference; the problem requirement transformation process includes problem-to-test requirement mapping, requirement standardization, and requirement structure generation; the task identification process includes identifying digital and physical tasks, analyzing the mapping relationship between digital and physical tasks to identify digital-physical fusion tasks, and if a mismatch is found between digital and physical tasks during the identification process of digital-physical fusion tasks, feedback correction of digital and physical tasks is provided.

5. The method according to claim 3, characterized in that, Test object prompts are used to characterize the system under test and its key components; performance indicator prompts are used to characterize the type of performance to be tested. Test condition prompts are used to characterize the cutting process conditions during test execution; Evaluation metric keywords are used to characterize the metrics of test results; Digital test task prompts are used to characterize the tasks of building and simulating digital models; Physical testing task prompts are used to characterize the methods, equipment, and data acquisition, processing, and analysis tasks of physical testing; data-real fusion task prompts are used to characterize the fusion tasks of physical and digital test objects, scenarios, processes, and results.

6. The method according to claim 3, characterized in that, The knowledge dimensions include at least: test objects, performance indicators, test conditions, digital testing, physical testing, and data-real fusion.

7. The method according to claim 6, characterized in that, Based on the aforementioned keyword set and a pre-built test knowledge base, a large language model is used to generate digital testing schemes, physical testing schemes, and data-real fusion schemes, including: Search the test knowledge base for knowledge entries containing test object prompts, performance indicator prompts, test condition prompts, evaluation indicator prompts, digital test task prompts, physical test task prompts, and digital-real fusion task prompts to obtain the corresponding digital test knowledge, physical test knowledge, and digital-real fusion test knowledge. The digital test knowledge is used as prompt words, combined with prompt words for test objects, performance indicators, test conditions, and evaluation indicators, and a large language model is used to generate a digital test plan. The physical test knowledge is used as prompt words, combined with prompt words for test objects, performance indicators, test conditions, and evaluation indicators, and a physical test plan is generated using a large language model. The aforementioned data-real fusion test knowledge is used as prompt words, combined with prompt words for test objects, performance indicators, test conditions, and evaluation indicators, to generate a data-real fusion test plan using a large language model.

8. The method according to claim 1, characterized in that, The test knowledge base is constructed in the following manner: Determine the corresponding prompt word template for each knowledge dimension; Using a large language model, multiple knowledge entries corresponding to the knowledge dimension are generated based on the prompt word template. These multiple knowledge entries constitute a set of knowledge units corresponding to the knowledge dimension. The relationships between knowledge unit sets are determined based on the relationships between knowledge items in different knowledge unit sets, wherein the relationships include at least one of the following: application relationship, dependency relationship, verification relationship, mapping relationship, and compensation relationship; A test knowledge base is constructed based on all sets of knowledge units and the relationships between them.

9. A system for developing a numerical-real fusion testing scheme for CNC machine tools based on a large language model, characterized in that, This includes a topic prompt word generation module and a test plan generation module; among them, The topic prompt generation module is configured to use a constructed test requirement analysis agent to generate a set of topic prompts based on the input test questions and multiple preset prompt dimensions. The test requirement analysis agent is built using the ReAct framework based on a large language model. The test scheme generation module is configured to generate digital test schemes, physical test schemes, and digital-real fusion schemes based on the topic set and the pre-built test knowledge base using a large language model, so as to obtain a complete set of test schemes. The test knowledge base is constructed based on multiple preset knowledge dimensions, and consists of multiple knowledge unit sets and the relationships between each knowledge unit set. One knowledge dimension corresponds to one knowledge unit set.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method as described in any one of claims 1-8.

Images