A method and system for modifying geometric features of a dialogue-driven mechanical structure
Through the dialogue-driven method, intelligent algorithm models are used to process user input, automatically generate design modification parameters, and update the three-dimensional model data of geometric features of mechanical structures in real time, solving the problems of time-consuming and labor-intensive and poor user interaction, and achieving efficient and intuitive Design modification process.
Patent Information
- Application Number
- CN202411141703.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-20
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-08-20
AI Technical Summary
The geometric feature design of traditional mechanical structures requires manual adjustment of parameters or the use of predefined algorithms, which is time-consuming and labor-intensive, poor user interaction, and the inability to directly modify the three-dimensional model data in real-time dialogue, resulting in low design efficiency.
Using dialogue-driven method, the database is constructed by discrete processing of the three-dimensional model data of mechanical structure geometric features, collecting mechanical knowledge and historical dialogue corpus, and using intelligent algorithm models for step-by-step fine-tuning training, obtaining and analyzing geometric feature modification information, and automatically generating design Modify the parameters, generate a new coordinate matrix and update the three-dimensional model data.
Real-time and interactive design modification of geometric features of mechanical structures is realized, which greatly simplifies the operation process, improves design efficiency and quality, and provides design engineers with a convenient and fast design experience.
Smart Images

Figure CN118657914B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent industrial design, and in particular to a dialogue-driven mechanical structure geometric feature modification method and system. Background Art
[0002] With the rapid development of artificial intelligence technology, mechanical design is increasingly moving towards intelligent, integrated and collaborative development. In the process of mechanical equipment design, it is often necessary to modify the geometric features of the mechanical structure according to the user's design modification requirements. Traditional geometric feature design requires design engineers to manually adjust parameters or use predefined algorithms to modify them. However, this method is often limited by the personal experience and intuition of design engineers, and the modification steps are frequent, time-consuming and laborious. In addition, traditional software interaction methods cannot realize real-time dialogue to directly modify the three-dimensional model data of the geometric features of the mechanical structure, resulting in poor user interactivity. Therefore, how to efficiently, quickly, and dynamically modify the geometric features of the mechanical structure in real time, thereby improving the work efficiency and design quality of design engineers, is an urgent problem that needs to be solved in the field of mechanical equipment design.
[0003] As can be seen from the above, the existing modification methods have the problems of being time-consuming and labor-intensive, lacking user interactivity, and being inefficient. Summary of the invention
[0004] In view of the above problems, the present invention proposes a dialogue-driven mechanical structure geometric feature modification method and system.
[0005] To achieve the purpose of the present invention, a method for modifying geometric features of a mechanical structure driven by a dialogue is provided, comprising the following steps:
[0006] S1: Discretization of 3D model data of mechanical structure geometric features;
[0007] S2: collecting a mechanical knowledge corpus and a historical dialogue corpus, constructing a dialogue database using the historical dialogue corpus, and performing step-by-step fine-tuning training on the intelligent algorithm model using the mechanical knowledge corpus and the constructed dialogue database;
[0008] S3: Acquire and analyze geometric feature modification information and automatically generate design modification parameters;
[0009] S4: obtaining a new coordinate matrix according to the design modification parameters, and obtaining discretized data of the modified three-dimensional model using the new coordinate matrix;
[0010] S5: Display and confirm the modified three-dimensional model discretization data.
[0011] Furthermore, in the step S1, the specific process of discretizing the three-dimensional model data of the geometric features of the mechanical structure is: the three-dimensional model data of the geometric features of the mechanical structure is gridded or point clouded by Matlab software to obtain discretized data, and the discretized data is divided into coordinate data and attribute data.
[0012] Furthermore, the coordinate data is a data matrix of grid coordinates or point coordinates extracted by gridding or point clouding of the three-dimensional model, and the attribute data is a data matrix including color, intensity, transparency, material and normal vector.
[0013] Furthermore, in step S2, the construction of the dialogue database is to establish a relationship database among historical geometric feature modification information, historical geometric feature modification results and preset prompt words based on the corpus data in the collected historical dialogue corpus, and the preset prompt words include the shape and size of the geometric features.
[0014] Furthermore, in step S2, the specific process of training the intelligent algorithm model by the step-by-step fine-tuning method includes: firstly performing preliminary training on the intelligent algorithm model using the mechanical knowledge corpus, and then retraining the preliminarily trained intelligent algorithm model using the dialogue database.
[0015] Furthermore, in step S3, the geometric feature modification information is obtained by processing voice or text information using a trained intelligent algorithm model.
[0016] Furthermore, in step S3, the steps of generating design modification parameters are as follows:
[0017] Establish a mapping relationship between historical geometric feature modification information and historical design modification parameters to form data annotation;
[0018] Importing the annotated historical geometric feature modification information and historical design modification parameters into a recurrent neural network algorithm model for training;
[0019] The geometric feature modification information obtained by analyzing the trained intelligent algorithm model is imported into the trained recurrent neural network algorithm model to intelligently generate design modification parameters.
[0020] Furthermore, in step S4, the new coordinate matrix is obtained by cross-producting the design modification parameters with the coordinate data, and the discretized data of the modified three-dimensional model is obtained by formatting and splicing the new coordinate matrix with the attribute data.
[0021] The present invention also provides a dialogue-driven mechanical structure geometric feature modification system, comprising: a three-dimensional data discretization processing module, a modification information acquisition and calculation processing module, a drive modification module, an output module and a data storage module;
[0022] The three-dimensional data discretization processing module is used to import the three-dimensional model data of the geometric features of the mechanical structure and discretize it to obtain discretized data of the three-dimensional model;
[0023] The modification information acquisition and calculation processing module is used to receive and analyze the user's design modification information, intelligently generate design modification parameters, and use the design modification parameters to generate a new coordinate matrix;
[0024] The driving modification module is used to drive the new coordinate matrix and the attribute matrix to perform formatting and splicing operations, thereby generating modified geometric feature three-dimensional model discretization data;
[0025] The output module is used to display and output the discretized data of the modified geometric feature three-dimensional model;
[0026] The data storage module is used to download and save the discretized data of the modified geometric feature three-dimensional model.
[0027] Compared with the prior art, the present invention has the following beneficial technical effects:
[0028] This solution uses an intelligent algorithm model to collect and process the user's design modification information in a guided dialogue question-and-answer manner, and generates new three-dimensional model discretized data based on the modification information. The user can view the modified data in real time and download and save it. Modifying the three-dimensional model of the mechanical structure's geometric features through dialogue greatly simplifies the mechanical structure's geometric feature modification operation process. Real-time dialogue and interactive dynamic modification enable a more efficient and intuitive mechanical structure geometric feature design modification process, providing design engineers with a convenient, fast, and immersive design experience. In addition, this solution not only improves the work efficiency and design quality of design engineers, but also lays the foundation for intelligent design in the field of mechanical design. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 is a flow chart of a method for modifying geometric features of a mechanical structure driven by a dialogue according to an embodiment;
[0030] Figure 2 is a schematic diagram of a module of a dialogue-driven mechanical structure geometric feature modification system according to an embodiment;
[0031] Figure 3 The flowchart is a detailed diagram of a method for modifying geometric features of a dialogue-driven mechanical structure according to an embodiment. DETAILED DESCRIPTION
[0032] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0033] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0034] refer to Figure 1 As shown, Figure 1 The flowchart of a method for modifying geometric features of a mechanical structure driven by a dialogue is shown in FIG. 1 , and the method for modifying geometric features of a mechanical structure driven by a dialogue comprises the following steps:
[0035] S1: Discretization of 3D model data of mechanical structure geometric features;
[0036] S2: collecting a mechanical knowledge corpus and a historical dialogue corpus, constructing a dialogue database using the historical dialogue corpus, and performing step-by-step fine-tuning training on the intelligent algorithm model using the mechanical knowledge corpus and the constructed dialogue database;
[0037] S3: Acquire and analyze geometric feature modification information and automatically generate design modification parameters;
[0038] S4: obtaining a new coordinate matrix according to the design modification parameters, and obtaining discretized data of the modified three-dimensional model using the new coordinate matrix;
[0039] S5: Display and confirm the modified three-dimensional model discretization data.
[0040] In one embodiment, in step S1, the specific process of discretizing the three-dimensional model data of the geometric features of the mechanical structure is: the three-dimensional model data of the geometric features of the mechanical structure is gridded or point clouded by Matlab software to obtain discretized data, and the discretized data is divided into coordinate data and attribute data.
[0041] In one embodiment, the coordinate data is a data matrix of grid coordinates or point coordinates extracted by gridding or point clouding the three-dimensional model, and the attribute data is a data matrix including color, intensity, transparency, material and normal vector.
[0042] In one embodiment, in step S2, the construction of the dialogue database is to establish a relationship database among historical geometric feature modification information, historical geometric feature modification results and preset prompt words based on the corpus data in the collected historical dialogue corpus, and the preset prompt words include the shape and size of the geometric features.
[0043] In one embodiment, in step S2, the specific process of training the intelligent algorithm model using the step-by-step fine-tuning method includes: first using the mechanical knowledge corpus to perform preliminary training on the intelligent algorithm model, and then using the dialogue database to retrain the preliminarily trained intelligent algorithm model.
[0044] In one embodiment, in step S3, the geometric feature modification information is obtained by processing voice or text information using a trained intelligent algorithm model.
[0045] In one embodiment, in step S3, the steps of generating design modification parameters are as follows:
[0046] Establish a mapping relationship between historical geometric feature modification information and historical design modification parameters to form data annotation;
[0047] Importing the annotated historical geometric feature modification information and historical design modification parameters into a recurrent neural network algorithm model for training;
[0048] The geometric feature modification information obtained by analyzing the trained intelligent algorithm model is imported into the trained recurrent neural network algorithm model to intelligently generate design modification parameters.
[0049] In one embodiment, in step S4, the new coordinate matrix is obtained by cross-producting the design modification parameters with the coordinate data, and the discretized data of the modified three-dimensional model is obtained by formatting and splicing the new coordinate matrix with the attribute data.
[0050] In one embodiment, Figure 2 As shown, a dialogue-driven mechanical structure geometric feature modification system is also provided, including: a three-dimensional data discretization processing module, a modification information acquisition and calculation processing module, a drive modification module, an output module and a data storage module;
[0051] The three-dimensional data discretization processing module is used to import the three-dimensional model data of the geometric features of the mechanical structure and discretize it to obtain discretized data of the three-dimensional model;
[0052] The modification information acquisition and calculation processing module is used to receive and analyze the user's design modification information, intelligently generate design modification parameters, and use the design modification parameters to generate a new coordinate matrix;
[0053] The driving modification module is used to drive the new coordinate matrix and the attribute matrix to perform formatting and splicing operations, thereby generating modified geometric feature three-dimensional model discretization data;
[0054] The output module is used to display and output the discretized data of the modified geometric feature three-dimensional model;
[0055] The data storage module is used to download and save the discretized data of the modified geometric feature three-dimensional model.
[0056] In one embodiment, Figure 3 The method for modifying geometric features of a mechanical structure driven by a dialogue comprises the following specific steps:
[0057] (1) Discretization of 3D model data of mechanical structure geometric features
[0058] Case of cuboid size modification. The three-dimensional model data of the cuboid is imported into Matlab (Matrix & Laboratory) software for discretization. First, the point cloud processing tool in the PCL function library of Matlab software is used to process the three-dimensional model data of the cuboid into point cloud data. The main steps include: cleaning, filtering and smoothing operations to obtain the discretized data of the cuboid, and the discretized data is divided according to regions to obtain coordinate data and attribute data; then the grid2points function in the python function library is used to perform matrix processing on the coordinate data and attribute data to obtain three-column data matrices. and the data matrix The three-column data matrix The three columns of data correspond to the X, Y, and Z coordinate directions of the discretized data. The data matrix Contains the color, intensity, transparency, material and normal vector of the discretized 3D model. Among them, the three-column data matrix and the data matrix The specific values are as follows:
[0059] ;
[0060] ;
[0061] (2) collecting a mechanical knowledge corpus and a historical dialogue corpus, constructing a dialogue database using the historical dialogue corpus, and using the mechanical knowledge corpus and the constructed dialogue database to perform step-by-step fine-tuning training on the intelligent algorithm model;
[0062] First, a large amount of knowledge on mechanical structure design is obtained through crawler methods, and a large amount of historical dialogue corpus is collected. Based on the large amount of mechanical design knowledge and historical dialogue corpus collected, a mechanical knowledge corpus and a historical dialogue corpus are formed. Among them, the mechanical knowledge corpus mainly contains basic definitions, theorems, formulas, industry planning expressions and other data in mechanical design, and the historical dialogue corpus mainly contains a large amount of geometric feature modification information and geometric feature modification result corpus. In order to better train the intelligent algorithm model, a dialogue database is constructed using the historical dialogue corpus. The dialogue database adds preset prompt words to the historical dialogue corpus, and establishes a relationship database between historical geometric feature modification information, historical geometric feature modification results and preset prompt words. The specific prompt word in this case uses [seg] as the prompt identifier, and the following examples are given:
[0063] Relationship data 1: "The volume of this cuboid is too large. Can it be reduced a little? It would be best if it could be reduced to 8 times of the original size: The volume of the cuboid is reduced to 8 times of the original size, and the corresponding length, width and height are reduced to 0.5 times of the original size respectively [seg] This is an example of modifying the size of a cuboid. The length, width and height correspond to the X, Y and Z coordinate directions respectively."
[0064] Relationship data 2: "The volume of this cuboid is too large. Can the volume of the cuboid be reduced a little? It would be best if it could be reduced to 8 times of the original size, and the X and Y directions should remain unchanged: the volume of the cuboid is reduced to 8 times of the original size, the X and Y directions remain unchanged, its length and width remain unchanged, and the height is reduced to 8 times of the original size [seg] This is an example of modifying the size of a cuboid. The length, width, and height correspond to the X, Y, and Z coordinate directions respectively."
[0065] The intelligent algorithm model used in this example is the general open source large language model LLaMA 2. In order to better train the LLaMA 2 large language model, LLaMA 2 is trained by step-by-step fine-tuning. The specific steps are as follows: First, the intelligent algorithm model is initially trained using the mechanical knowledge corpus to enhance the basic semantic understanding ability of the LLaMA2 model in the field of mechanical design; then the preliminarily trained large language model is retrained using the dialogue corpus, which not only makes the model converge quickly, but also enhances the guided question-answering ability in the field of mechanical design and better parses geometric feature modification information. Finally, the trained intelligent algorithm model is saved.
[0066] (3) Obtain and analyze geometric feature modification information and automatically generate design modification parameters
[0067] Intelligently obtain design modification information. Use a human-computer interaction interface to receive text or voice information input by the user. In this embodiment, the user inputs the modification text information as follows: "The volume of this cuboid is relatively large, and the size of the cuboid needs to be reduced by 8 times", and imports the geometric feature modification information into the above-mentioned trained intelligent algorithm model.
[0068] Intelligent analysis of design modification information. First, the modified information entered by the user is semantically analyzed using the trained large language model and output in a standardized format. For example, the geometric feature modification information entered by the user is: the volume of this cuboid is too large, and the size of the cuboid needs to be reduced by 8 times. The intelligent algorithm model extracts features from this modification information and performs analysis, and outputs the analysis results: the volume of the cuboid is reduced by 8 times, and its length, width, and height are all reduced to 0.5 times the original.
[0069] Intelligently generate design modification parameters. First, collect a large amount of historical geometric feature modification information and historical design modification parameters, where the historical geometric feature modification information is mainly the size information of the geometric features; then, establish a mapping relationship between the historical geometric feature modification information and the historical design modification parameters to form data annotation.
[0070] Examples of historical geometry feature modification information and historical design modification parameters for this case are as follows:
[0071] Example 1: Historical geometric feature modification information: The volume of the cuboid is reduced to 8 times the original, and the corresponding length, width and height are reduced to 0.5 times the original. Historical design modification parameters: X direction (length): 0.5, Y direction (width): 0.5, Z direction (height): 0.5
[0072] Example 2: Historical geometric feature modification information: The volume of the cuboid is reduced to 8 times the original, the X and Y directions remain unchanged, its length and width remain unchanged, and the height is reduced to 8 times the original. Historical design modification parameters: Z direction (height): 0.125
[0073] Then, the labeled data is imported into the RNN (Recurrent Neural Network) algorithm model for training, and the trained RNN algorithm model is saved. The design modification information analyzed by the large language model after step-by-step fine-tuning (the volume of the rectangular block is reduced by 8 times, and its length, width, and height are all reduced to 0.5 times the original) is imported into the trained RNN algorithm model. After calculation and processing by the model, the corresponding design modification parameters are generated: X direction (length): 0.5, Y direction (width): 0.5, Z direction (height): 0.5.
[0074] (4) obtaining a new coordinate matrix according to the design modification parameters, and obtaining discretized data of the modified three-dimensional model using the new coordinate matrix;
[0075] First, the generated design modification parameters are converted into a data matrix K using a program. The specific values of the data matrix K in this case are as follows:
[0076] ;
[0077] Then, the data matrix obtained above is With the three-column data matrix Perform a cross multiplication operation to obtain a new coordinate matrix The specific calculation process is as follows:
[0078] ;
[0079] Finally, in this embodiment, the new coordinate matrix With the data matrix Perform left and right splicing operations to obtain the discretized data matrix of the modified three-dimensional model The specific calculation steps are as follows:
[0080] ;
[0081] In addition, the grid2points function in the python function library is used to transform the matrix Processing is performed to obtain discretized data of the modified three-dimensional model.
[0082] (5) Display and confirm the modified 3D model discretization data
[0083] In this case, the discretization data display tool CloudCompare is used to display the new 3D model discretization data in real time. The user confirms whether the modified data meets the requirements based on the real-time display results. If the confirmation is passed, the user can download and save the data. If the user fails to confirm, the above steps (3), (4), and (5) are executed again until the user confirms that it is passed.
[0084] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0085] It should be noted that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects, and do not represent a specific order for the objects. It is understandable that the specific order or sequence of "first\second\third" can be interchanged where permitted. It should be understood that the objects distinguished by "first\second\third" can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.
[0086] The terms "including" and "having" and any variations thereof in the embodiments of the present application are intended to cover non-exclusive inclusions. For example, a process, method, device, product, or equipment that includes a series of steps or modules is not limited to the listed steps or modules, but may optionally include steps or modules that are not listed, or may optionally include other steps or modules that are inherent to these processes, methods, products, or equipment.
[0087] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.
Claims
1. A method for modifying geometric features of a mechanical structure driven by a dialogue, characterized in that: The steps include: S1: Discretization of 3D model data of mechanical structure geometric features; S2: collecting a mechanical knowledge corpus and a historical dialogue corpus, constructing a dialogue database using the historical dialogue corpus, and performing step-by-step fine-tuning training on the intelligent algorithm model using the mechanical knowledge corpus and the constructed dialogue database; S3: using a human-computer interaction interface to collect user input information; then using the intelligent algorithm model that has been trained through step-by-step fine-tuning to perform an analysis operation on the user input information, and obtain geometric feature modification information; Then, a recurrent neural network algorithm model is built and trained; finally, a design modification parameter is generated based on the geometric feature modification information and the trained recurrent neural network algorithm model; S4: obtaining a new coordinate matrix according to the design modification parameters, and obtaining discretized data of the modified three-dimensional model using the new coordinate matrix; S5: Displaying and confirming the modified three-dimensional model discretization data.
2. The method for modifying geometric features of a mechanical structure driven by a dialogue according to claim 1, characterized in that: In step S1, the specific process of discretizing the three-dimensional model data of the geometric features of the mechanical structure is: the three-dimensional model data of the geometric features of the mechanical structure is gridded or point clouded by Matlab software to obtain discretized data, and the discretized data is divided into coordinate data and attribute data.
3. The method for modifying geometric features of a mechanical structure driven by a dialogue according to claim 2, characterized in that: The coordinate data is a data matrix of grid coordinates or point coordinates extracted by gridding or point clouding of the three-dimensional model, and the attribute data is a data matrix including color, intensity, transparency, material and normal vector.
4. The method for modifying geometric features of a mechanical structure driven by a dialogue according to claim 1, characterized in that: In step S2, the dialogue database is constructed by establishing a relationship database among historical geometric feature modification information, historical geometric feature modification results and preset prompt words based on the corpus data in the collected historical dialogue corpus, wherein the preset prompt words include the shape and size of the geometric features.
5. The method for modifying geometric features of a mechanical structure driven by a dialogue according to claim 1, characterized in that: In step S2, the specific process of training the intelligent algorithm model by the step-by-step fine-tuning method includes: firstly using the mechanical knowledge corpus to perform preliminary training on the intelligent algorithm model, and then using the dialogue database to retrain the preliminarily trained intelligent algorithm model.
6. The method for modifying geometric features of a mechanical structure driven by a dialogue according to claim 4, characterized in that: In step S3, the steps of generating design modification parameters are as follows: Establishing a mapping relationship between the historical geometric feature modification information and the historical design modification parameters to form data annotation; Importing the annotated historical geometric feature modification information and historical design modification parameters into a recurrent neural network algorithm model for training; The geometric feature modification information obtained by analyzing the trained intelligent algorithm model is imported into the trained recurrent neural network algorithm model to intelligently generate design modification parameters.
7. The method for modifying geometric features of a mechanical structure driven by a dialogue according to claim 3, characterized in that: In step S4, the new coordinate matrix is obtained by cross-producting the design modification parameters with the coordinate data, and the discretized data of the modified three-dimensional model is obtained by formatting and splicing the new coordinate matrix with the attribute data.
8. A dialogue-driven mechanical structure geometric feature modification system, characterized in that: include: Three-dimensional data discretization processing module, modification information acquisition and calculation processing module, drive modification module, output module and data storage module; The three-dimensional data discretization processing module is used to import the three-dimensional model data of the geometric features of the mechanical structure and discretize it to obtain discretized data of the three-dimensional model; The modification information acquisition and calculation processing module is used to receive user input information, and after processing the user input information, generate design modification parameters, and then use the design modification parameters to generate a new coordinate matrix; The driving modification module is used to drive the new coordinate matrix and the attribute matrix to perform formatting and splicing operations, thereby generating discretized data of the modified geometric feature three-dimensional model; The output module is used to display and output the discretized data of the modified geometric feature three-dimensional model; The data storage module is used to download and save the discretized data of the modified geometric feature three-dimensional model.
