Intelligent modeling method and device, electronic equipment and storage medium

Abstract

The invention provides an intelligent modeling method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: inputting modeling demand information into a first large model, and outputting modeling logic information; the modeling logic information is input into the second large model, and modeling process information and modeling detailed operation information are output; modeling process information and modeling detailed operation information are input into the third model, and a modeling language is output; generating a CAD model based on the modeling language; wherein the first large model is obtained by training based on a CAD model carrying a modeling sequence and modeling specification information; the second large model is obtained by training based on modeling software course information; and the third model is obtained based on programming grammar and code library training required by modeling. According to the method, manual modeling is replaced with a large model, automatic and rapid creation of the model is achieved, manual modeling time is shortened, and meanwhile the efficiency and quality of modeling design work are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an intelligent modeling method, device, electronic device and storage medium. Background Art

[0002] As the functions of modeling and design software continue to expand, its operation becomes more and more complicated. Designers need to spend a lot of time learning and mastering these complex operation processes and tool usage methods. Since a lot of time is spent on learning software operation, the time and energy that designers actually use for creative ideas and actual design work are dispersed. In addition, during the operation process, operational errors may occur due to the complexity of the software, which further increases the difficulty of the design work. At the same time, the complex software operation requirements make the design work higher The threshold for personnel's software operation ability. It can be seen that the use of manual modeling will affect the efficiency of modeling design. Summary of the invention

[0003] The present invention provides an intelligent modeling method, device, electronic device and storage medium to solve the defect of low efficiency of manual modeling in the prior art, realize automatic modeling based on large models, reduce manual modeling time, and improve the efficiency and quality of modeling design work.

[0004] The present invention provides an intelligent modeling method, comprising the following steps: Inputting modeling requirement information into the first large model to obtain modeling logic information output by the first large model; Inputting the modeling logic information into the second largest model to obtain modeling process information and modeling detailed operation information output by the second largest model; Inputting the modeling process information and the modeling detailed operation information into a third large model to obtain a modeling language output by the third large model; Based on the modeling language, a computer-aided design (CAD) model is generated; Among them, the first large model is trained based on the CAD model carrying the modeling sequence and modeling specification information; the second large model is trained based on the modeling software tutorial information; and the third large model is trained based on the programming syntax and code library required for modeling.

[0005] According to an intelligent modeling method provided by the present invention, after inputting the modeling requirement information into the first large model and obtaining the modeling logic information output by the first large model, the method further includes: Performing a first check on the modeling logic information; If the first verification fails, modifying the modeling logic information based on the first verification result; The modified modeling logic information is input into the second large model to obtain the modeling process information and modeling detailed operation information output by the second large model.

[0006] According to an intelligent modeling method provided by the present invention, after inputting the modeling logic information into the second large model and obtaining the modeling process information and modeling detailed operation information output by the second large model, the method further includes: Performing a second verification on the modeling process information and the modeling detailed operation information; If the second verification fails, modifying the modeling process information and the modeling detailed operation information based on the second verification result; The modified modeling process information and the modified modeling detailed operation information are input into the third largest model to obtain the modeling language output by the third largest model.

[0007] According to an intelligent modeling method provided by the present invention, after inputting the modeling requirement information into the first large model and obtaining the modeling logic information output by the first large model, the method further includes: Matching parts from a parts library according to the modeling logic information; the parts library includes modeling information of the CAD models that have been completed; When the target part is matched, the modeling process information, detailed modeling operation information and modeling language of the target part are obtained from the parts library; According to the difference information between the preset modeling requirements and the target part, fine-tune the modeling process information, the modeling detailed operation information and the modeling language of the target part to obtain the target modeling process information, the target modeling detailed operation information and the target modeling language; The target modeling process information and the target modeling detailed operation information are sent to the second large model, and the target modeling language is sent to the third large model to generate the CAD model.

[0008] According to an intelligent modeling method provided by the present invention, the first large model includes a conversion layer; the step of inputting modeling requirement information into the first large model to obtain modeling logic information output by the first large model includes: The modeling requirement text is input into the conversion layer for modeling logic conversion to obtain the modeling logic information output by the conversion layer.

[0009] According to an intelligent modeling method provided by the present invention, the second large model includes a decomposition layer and a conversion layer; the step of inputting the modeling logic information into the second large model to obtain the modeling process information and modeling detailed operation information output by the second large model includes: Inputting the modeling logic information into the decomposition layer to decompose the modeling process, and obtaining the modeling process information output by the decomposition layer; The modeling process information is input into the conversion layer for operation information conversion, and the modeling detailed operation information of each step of the modeling process output by the conversion layer is obtained.

[0010] According to an intelligent modeling method provided by the present invention, the third model includes a coding layer; the step of inputting the modeling process information and the modeling detailed operation information into the third model to obtain a modeling language output by the third model includes: The modeling process information and the modeling detailed operation information are input into the encoding layer for automatic encoding to obtain the modeling language output by the encoding layer.

[0011] The present invention also provides an intelligent modeling device, comprising the following modules: A first modeling module, used for inputting modeling requirement information into a first large model to obtain modeling logic information output by the first large model; A second modeling module, used for inputting the modeling logic information into a second large model, and obtaining modeling process information and modeling detailed operation information output by the second large model; A third modeling module, used for inputting the modeling process information and the modeling detailed operation information into a third large model to obtain a modeling language output by the third large model; The CAD model generation module is used to generate a computer-aided design CAD model based on the modeling language; wherein the first large model is trained based on a CAD model carrying a modeling sequence and modeling specification information; the second large model is trained based on modeling software tutorial information; and the third large model is trained based on the programming syntax and code library required for modeling.

[0012] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the intelligent modeling method described above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the intelligent modeling methods described above.

[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the intelligent modeling method described above is implemented.

[0015] The intelligent modeling method, device, electronic device and storage medium provided by the present invention input the modeling requirement information into the first large model to obtain the modeling logic information output by the first large model; input the modeling logic information into the second large model to obtain the modeling process information and modeling detailed operation information output by the second large model; input the modeling process information and modeling detailed operation information into the third large model to obtain the modeling language output by the third large model; based on the modeling language, a computer-aided design CAD model is generated; wherein the first large model is obtained by training based on the CAD model carrying the modeling sequence and the modeling specification information; the second large model is obtained by training based on the modeling software tutorial information; and the third large model is obtained by training based on the programming grammar and code library required for modeling. The present invention uses a large model to replace manual modeling, realizes the automation and rapid creation of the model, reduces the manual modeling time, and improves the efficiency and quality of the modeling design work. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of the intelligent modeling method provided by the present invention.

[0018] Figure 2 It is a flow chart of the intelligent modeling method based on a large model provided by the present invention.

[0019] Figure 3 It is a structural schematic diagram of the intelligent modeling device provided by the present invention.

[0020] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] It should be noted that in the early stages of product design, designers often have active minds, and various novel ideas and different structural solutions will continue to emerge in their minds. However, these ideas that only stay at the level of ideas need to be modeled with the help of CAD (Computer Aided Design) software to transform them into specific, visual models, so as to verify the feasibility of these ideas and solutions in practice. However, manually using CAD software to realize these creative models is a very time-consuming process, because building a complete model requires many steps, such as accurately drawing the geometric shapes of each component, accurately setting the dimensions and constraints between them, etc. In the long process of manual modeling, due to the cumbersome operation and the possibility of repeated modifications, the designer's original original ideas are likely to be deformed or even lost in the process. In short, the speed of model realization cannot keep up with the speed of creativity, which limits the development of innovation. If designers can quickly transform ideas into specific solutions through CAD software to verify their feasibility, the speed and quality of innovation will be improved.

[0023] In the related art, artificial intelligence technology has made significant progress, but in the field of mechanical design and modeling, the application of artificial intelligence is relatively limited. The present invention considers using large-scale artificial intelligence models to replace manual modeling to achieve automation and rapid creation of models. By simplifying software operations, designers can focus more on the generation and conception of ideas, ensuring the integrity and accuracy of ideas in the implementation process, rather than being disturbed by complex software operations, so that designers can devote more energy to the discovery and realization of ideas, and improve the efficiency and quality of design work.

[0024] The present invention provides an intelligent modeling method, aiming to solve the problems of high learning cost, low efficiency and low quality of manual modeling design caused by the complexity of design software operation.

[0025] Combine the following Figure 1-Figure 4 The intelligent modeling method, device, electronic device and storage medium of the present invention are described.

[0026] Figure 1 It is a flow chart of the intelligent modeling method provided by the present invention, such as Figure 1 As shown, the method includes the following: Step 101: input modeling requirement information into a first large model to obtain modeling logic information output by the first large model.

[0027] The modeling requirement information may include modeling requirement text and modeling requirements input by voice.

[0028] The first large model is trained based on CAD models with modeling order and modeling specification information. It is understandable that the CAD model itself is the result of digital design presentation of products or structures, etc., which includes many elements such as geometric shapes, dimensioning, and constraint relationships; and the CAD model with modeling order means that these models not only show the final design results, but also record in detail the sequence of each step of building the model from scratch. For example, for a CAD model of a mechanical part, it clearly presents a series of orderly modeling processes such as first creating a basic component in the shape of a disk through stretching operations, and then punching holes on this disk. This CAD model with order provides a very intuitive and specific modeling practice example for the first large model, enabling the first large model to directly observe the actual modeling process.

[0029] Modeling specification information specifies the standards and requirements that should be followed in modeling from a more macro and principled perspective, covering information such as the order of different operations, dimensional accuracy control standards, and the matching relationship criteria between components. For the first large model, the modeling specification provides it with a clear framework and rule system, so that it understands that the modeling sequence and operation methods are correct and reasonable.

[0030] The first model summarizes the general modeling rules and sequences by studying a large number of CAD models with modeling sequences and corresponding modeling specifications. After learning many specific modeling sequences and specifications, the scattered knowledge and experience are integrated to form a set of modeling logic. When faced with new modeling requirements, the modeling requirements can be converted into modeling logic.

[0031] In one embodiment, the first large model can be trained based on the following steps: Step 1: Data collection and preprocessing stage: 1.1) Data collection: Collect CAD models with modeling sequence and modeling specification information; 1.2) Data cleaning and formatting: Clean the collected CAD model data with modeling sequence to remove possible erroneous information in the model (such as invalid geometric elements, incorrect constraint relationships, etc.). At the same time, unify the data format of the CAD model so that it can be effectively read and processed by the model training system; for the collected modeling specification information, organize the text content, extract key rules and key points, and convert it into a machine-readable format.

[0032] 1.3) Data annotation (optional): If the modeling order in the CAD model is not obvious or the learning focus needs to be further strengthened, the model can be annotated. The annotation content can include the name, purpose, and relationship with the previous and next steps of each modeling step. In this way, during the training process, the model can more clearly understand the meaning and order of each step.

[0033] Step 2: Data collection and preprocessing stage: 2.1) Model architecture selection and initialization phase: Select a suitable neural network architecture as the basis of the first model according to the characteristics of the task. For example, if the focus is on learning sequential information, you can choose a recurrent neural network (RNN) or its variants (such as long short-term memory networks or gated recurrent units) because they can effectively process sequence data, and modeling sequence is essentially a kind of sequence information. Alternatively, you can use the Transformer architecture, whose self-attention mechanism can well capture the relationship between different modeling steps.

[0034] 2.2) Initialize model parameters: After selecting the architecture, initialize the model parameters. For example, use random initialization to assign initial values ​​to the weights and biases of the neural network.

[0035] Step 3: Training phase: 3.1) Input data processing: The pre-processed CAD model with modeling sequence and modeling specification information is input into the model.

[0036] 3.2) Forward propagation: Data is propagated forward through the layers of the neural network. In this process, the model calculates the input data according to the current parameter settings and generates a prediction result.

[0037] 3.3) Calculate the loss function: Compare the model’s prediction results with the true modeling order (the correct order extracted from the CAD model data) and calculate the loss function.

[0038] 3.4) Back propagation and parameter update: Based on the calculated loss function, the model parameters are updated through the back propagation algorithm.

[0039] 3.5) Multiple Iterations of Training: Repeat the above process of forward propagation, loss function calculation, back propagation and parameter update, and train multiple times until the loss of the model on the training data reaches an acceptable level or no longer decreases significantly.

[0040] Step 4: Model evaluation and verification phase.

[0041] Step 5: Model testing phase.

[0042] After the first large model is obtained through training, the modeling requirement information is input into the first large model to obtain the modeling logic information output by the first large model. For example, through human-computer interaction, the modeling requirements expressed by the designer in voice or text form are obtained, and the modeling requirements are converted into modeling logic information based on the first large model, and the modeling logic information includes the modeling process.

[0043] In one embodiment, the first large model includes a conversion layer, and the modeling requirement text is input into the conversion layer for modeling logic conversion to obtain modeling logic information output by the conversion layer, and the modeling logic information is editable. The conversion layer operates through a series of internal mechanisms and algorithms to convert the relatively abstract and vague modeling requirements presented in text form into clear, specific modeling logic information that includes a complete modeling process.

[0044] For example, the conversion layer first receives modeling requirements expressed in text from designers or other relevant personnel. These requirements may cover various requirements for products or projects, such as "designing an object similar to a sphere but with irregular protrusions on the surface" in terms of shape, "the object must be able to achieve telescopic function" in terms of function, and "the surface must have specific colors and textures" in terms of appearance details. After receiving the modeling requirement text, the conversion layer uses natural language processing technology and algorithms to deeply understand the true meaning and intention of the text to identify the key elements. For example, for the above example, it identifies key descriptions such as "sphere", "irregular protrusions", "telescopic function", "specific color", "texture", and parses the relationship between them to determine whether these requirements are met at the same time or in a certain order. Based on the accurate understanding of the modeling requirement text, the conversion layer converts the modeling requirements into modeling logic.

[0045] Step 102: input the modeling logic information into the second largest model to obtain the modeling process information and modeling detailed operation information output by the second largest model.

[0046] The second largest model is trained based on the modeling software tutorial information. Among them, the modeling software tutorial information can be a CAD software tutorial, which details the various functions of the CAD software, how to use the tools, the steps for building different types of models, and related precautions. For example, for a CAD modeling tutorial for a mechanical part, it will explain how to draw basic geometric figures, how to perform operations such as stretching and rotation to form the shape of the part, and how to add dimensioning and constraints. The second largest model uses these CAD software tutorials as a source of training data, aiming to deeply learn the usage rules of CAD software and knowledge related to the modeling process. The second largest model simulates the learning process by studying the CAD software tutorial, so that it can better master the internal operating mechanism and modeling logic of the CAD software.

[0047] In one embodiment, the second largest model can be trained based on the following steps: Step 1: Data collection and preprocessing stage: 1.1) Collect CAD software tutorial data: CAD software tutorials are collected from various channels, including official documents, online tutorial websites, professional books, etc. These tutorials cover various functional modules of CAD software, such as the use of basic drawing tools (straight lines, circles, polygons, etc.), 3D modeling operations (stretching, rotating, lofting, etc.), model editing (trim, chamfering, Boolean operations, etc.) and advanced functions (parametric design, assembly design, etc.).

[0048] 1.2) Data cleaning and formatting: Clean the collected tutorial texts to remove irrelevant advertisements, incorrect formats (such as garbled characters, irregular punctuation, etc.) and repeated content. At the same time, unify the text format, such as standardizing the chapter titles and step numbers in different tutorials, so that the model can more easily understand the structure and content of the tutorials.

[0049] 1.3) Data annotation (optional): Based on the training objectives, it is helpful to annotate the key content in the tutorial. For example, annotate the CAD software function corresponding to each modeling operation, the sequence of operations (such as "first step", "second step", etc.), the goal of the operation (such as "create the main shape", "add detail features"), etc. These annotations can provide clearer learning clues for the model.

[0050] Step 2: Model architecture selection and initialization phase: 2.1) Choose a suitable neural network architecture: Transformer architecture or its variants: Since the CAD software tutorial text has a certain sequence (there is a sequence between steps), the self-attention mechanism of the Transformer architecture can handle this sequence information well, and can focus on the relationship between different parts of the tutorial, such as the association between an operation step and its previous and subsequent steps, and the connection between the software functions involved in different operations. Combined with Convolutional Neural Networks (CNN): Considering that the tutorial text may contain some auxiliary information such as graphics and tables, CNN can be partially combined to extract features from these non-text information. For example, extract features such as software tool icons and menu layouts from the screenshots of the operation interface in the tutorial, and process them together with the text information.

[0051] 2.2) Initialize model parameters: Initialize the parameters of the model according to the requirements of the selected architecture.

[0052] Step 3: Training phase: 3.1) Input data processing: Input the preprocessed CAD software tutorial data into the model. For the text part, the text can be converted into machine-readable forms such as word vectors or character vectors. For example, a pre-trained word embedding model (such as Word2Vec, Bert, etc.) can be used to map the vocabulary in the tutorial to a low-dimensional vector space. If there is annotation information, it is also used as input together with the text vector. For auxiliary information such as graphics that may exist, after feature extraction such as CNN, its feature vector is fused with the text vector.

[0053] 3.2) Forward propagation: Data is propagated forward through the layers of the neural network. In this process, the model calculates the input data according to the current parameter settings and generates a prediction result.

[0054] 3.3) Calculate the loss function: Compare the model’s prediction results with the actual information in the tutorial (such as correct operation steps, software function usage, etc.) and calculate the loss function.

[0055] 3.4) Back propagation and parameter update: Based on the calculated loss function, the model parameters are updated through the back propagation algorithm.

[0056] 3.5) Multiple Iterations of Training: Repeat the above process of forward propagation, loss function calculation, back propagation and parameter update, and train multiple times until the loss of the model on the training data reaches an acceptable level or no longer decreases significantly.

[0057] Step 4: Model evaluation and verification phase.

[0058] Step 5: Model testing phase.

[0059] After the training of the second largest model is completed, the modeling logic information is input into the second largest model to obtain the modeling process information and modeling detailed operation information output by the second largest model, wherein the modeling process information and modeling detailed operation information are both editable. For example, the second largest model uses its language understanding ability to decompose and convert the modeling process formed by the first largest model into the corresponding required feature combination and the detailed operations corresponding to each feature. Assume that a preliminary modeling process is formed through the first largest model (for example, the modeling process of designing a simple mechanical device, first creating the main frame, and then adding some auxiliary components, etc.). The second largest model will decompose this modeling process and split it into more detailed steps and links. For example, for the step of creating the main frame, it is further decomposed into more specific sub-steps such as drawing basic geometric figures (such as rectangles, circles, etc.), converting these geometric figures into solid parts through operations such as stretching or rotation, and setting constraints between parts. After decomposition, the second largest model will convert these detailed steps into the corresponding required feature combination and the detailed operations corresponding to each feature. Each sub-step has its corresponding features. For example, for the sub-step of drawing a rectangle, the features may include the side length and position coordinates of the rectangle; the features of the stretching operation may include the stretching direction and height.

[0060] In one embodiment, the second large model includes a decomposition layer and a conversion layer. The modeling logic information is input into the decomposition layer to decompose the modeling process, and the modeling process information output by the decomposition layer is obtained; the modeling process information is input into the conversion layer to convert the operation information, and the modeling detailed operation information of each step of the modeling process is obtained by the output of the conversion layer.

[0061] The main function of the decomposition layer is to decompose relatively abstract and general modeling logic information into more detailed modeling process information. The core function of the conversion layer is to convert the modeling process information output by the decomposition layer into specific and executable detailed operation information of each step of the modeling process.

[0062] For example, suppose the input modeling logic information is: design a cylindrical mechanical part, first create the cylindrical body, then punch holes in the cylinder, and finally perform surface polishing. The decomposition operations of the decomposition layer: create the cylindrical body, stretch it into a cylinder, punch holes in the cylinder, and perform surface polishing. Through the processing of the decomposition layer, the originally more general modeling logic information is refined into modeling process information containing specific sub-steps and operation sequences. The conversion work of the conversion layer is explained by taking the creation of the cylindrical body as an example: 1) Find the "Drawing" option in the main menu bar and click to expand the drop-down menu. The icon of the "Circle" tool is usually a circle. When you hover the mouse over the icon, a "Circle" prompt will be displayed. Click the icon to activate the circle drawing tool; 2) If you want to determine the center position of the circle by entering coordinate values, enter the specific coordinate values ​​(such as "X=10, Y=20" mentioned above) in the command line input area at the bottom of the software according to the design requirements, and then press the Enter key; 3) Enter the radius value of the circle (such as "R=5") in the command line input area, and then press the Enter key to complete the bottom circle.

[0063] Step 103: input the modeling process information and the modeling detailed operation information into the third largest model to obtain the modeling language output by the third largest model.

[0064] The third model is trained based on the programming syntax and code library required for modeling. It is understandable that during the operation of CAD software, the realization of its various functions is supported by the background code. These background codes have specific programming syntax and specify how to describe and implement various operations such as drawing, modeling, editing, etc. through code. The third model uses the programming syntax of these CAD background codes and related code libraries as the source of training data. The purpose is to deeply understand the operating mechanism at the internal code level of CAD software and master the implementation of various modeling operations through code. By learning this corresponding rule, the third model can accurately convert specific modeling operation descriptions into corresponding background codes when receiving them.

[0065] In one embodiment, the third largest model can be trained based on the following steps: Step 1: Data collection and preprocessing stage: 1.1) Collect CAD background code and code library: Collect CAD background code related information from channels such as the official documentation of the CAD software, developer resource websites, and code samples in the software installation directory.

[0066] 1.2) Data cleaning and formatting: Clean the collected code, remove comments, error messages (such as debugging code with syntax errors) and irrelevant code snippets (such as code used for testing but irrelevant to actual modeling operations). At the same time, unify the code format to make it conform to standard programming specifications and CAD software code style requirements.

[0067] 1.3) Data Annotation (Optional): Depending on your training goals, it may be helpful to annotate your code.

[0068] Step 2: Model architecture selection and initialization phase: 2.1) Choose the right neural network architecture: Sequence to Sequence (Seq2Seq) architecture, since the task is to convert detailed modeling operation descriptions into CAD software code, which is essentially a sequence to sequence conversion problem. The Seq2Seq architecture can handle this type of task well.

[0069] Transformer architecture or its variants: The self-attention mechanism of the Transformer architecture performs well in processing sequence data. It can effectively capture the relationship between elements in the input sequence (modeling operation description) and the output sequence (CAD software code).

[0070] 2.2) Initialize model parameters: Initialize the parameters of the model according to the requirements of the selected architecture.

[0071] Step 3: Training phase: 3.1) Input data processing: input the preprocessed CAD background code and modeling operation description data into the model.

[0072] 3.2) Forward propagation: Data is propagated forward through the layers of the neural network. In this process, the model calculates the input data according to the current parameter settings and generates a prediction result.

[0073] 3.3) Calculate the loss function: Compare the model’s predictions with the actual CAD software code (obtained from the collected and preprocessed data) and calculate the loss function.

[0074] 3.4) Back propagation and parameter update: Based on the calculated loss function, the model parameters are updated through the back propagation algorithm.

[0075] 3.5) Multiple Iterations of Training: Repeat the above process of forward propagation, loss function calculation, back propagation and parameter update, and train multiple times until the loss of the model on the training data reaches an acceptable level or no longer decreases significantly.

[0076] Step 4: Model evaluation and verification phase.

[0077] Step 5: Model testing phase.

[0078] After the third model is trained, the modeling process information and the modeling detailed operation information are input into the third model to obtain the modeling language output by the third model, wherein the modeling language is editable. In one embodiment, the third model includes a coding layer, and the modeling process information and the modeling detailed operation information are input into the coding layer for automatic coding to obtain the modeling language output by the coding layer. The coding layer converts the modeling process information and the modeling detailed operation information into codes recognizable by the CAD software.

[0079] For example, the coding layer must first understand the semantics of the input modeling process information and modeling detailed operation information. After understanding the semantics, the coding layer converts these semantics into specific code instructions. This involves converting abstract operations (such as punching holes at specific locations) into actual function calls, parameter settings, and other code forms in CAD software. For example, converting a punching operation into a code instruction like "hole-function (x, y, z, diameter, depth)", where (x, y, z) is the coordinate of the punching position, diameter is the diameter of the hole, and depth is the depth of the hole, and ensure that the settings of these parameters comply with the code specifications and syntax requirements of the CAD software.

[0080] Optionally, the modeling language may be verified. If the verification fails, the modeling language is modified based on the verification result, and then a CAD model is generated based on the modified modeling language.

[0081] By converting the modeling process information and modeling detailed operation information into code, the CAD modeling process can be automated. Once the corresponding code is generated, it can be directly input into the CAD software, and the software will automatically complete the modeling task according to the operation content and sequence specified by the code, greatly improving the efficiency and accuracy of modeling.

[0082] Step 104: Generate a computer-aided design (CAD) model based on the modeling language.

[0083] Through CAD secondary development technology, a CAD interface is developed to import the code from the previous step. The CAD software completes the modeling by receiving the code input from the previous step and using the modeling function of the software itself.

[0084] In one embodiment, the modeling requirement text is input into the large model to obtain the modeling language output by the large model; wherein the large model integrates the first large model, the second large model and the third large model.

[0085] The intelligent modeling method provided by the embodiment of the present invention inputs modeling requirement information into the first large model to obtain modeling logic information output by the first large model; inputs modeling logic information into the second large model to obtain modeling process information and modeling detailed operation information output by the second large model; inputs modeling process information and modeling detailed operation information into the third large model to obtain modeling language output by the third large model; based on the modeling language, a computer-aided design CAD model is generated; wherein the first large model is obtained by training based on a CAD model carrying a modeling sequence and modeling specification information; the second large model is obtained by training based on modeling software tutorial information; and the third large model is obtained by training based on programming syntax and code base required for modeling. The present invention uses a large model to replace manual modeling, realizes the automation and rapid creation of models, reduces the time of manual modeling, and improves the efficiency and quality of modeling design work.

[0086] Based on the above embodiment, after inputting the modeling requirement information into the first large model and obtaining the modeling logic information output by the first large model, the method further includes: Step 1010, performing a first check on the modeling logic information; Step 1011, when the first verification fails, modify the modeling logic information based on the first verification result; Step 1012: input the modified modeling logic information into the second large model to obtain the modeling process information and modeling detailed operation information output by the second large model.

[0087] It is understandable that the purpose of verification is to ensure that the modeling logic information output by the first model is reasonable and feasible in practical applications. The modeling logic information includes the modeling process and related logical elements. If there are errors or unreasonableness in it, such as the order of steps in the modeling process is reversed, or the technical parameters required for a certain step are set unreasonably, it will lead to failure in the actual modeling process or obtain a model that does not meet the requirements. Verification can confirm whether the modeling logic information truly meets the original modeling requirements.

[0088] A series of rules can be formulated for verification, which can be based on industry standards, engineering specifications, and previous modeling experience. When the verification fails, the verification results are used to modify the modeling logic information. For example, if it is a modeling process problem, the steps are rearranged in the correct order. Finally, the modified modeling logic information is input into the second largest model to obtain the modeling process information and modeling detailed operation information output by the second largest model.

[0089] The embodiments of the present invention can ensure the accuracy of modeling logic information through verification, thereby improving modeling accuracy.

[0090] Based on the above embodiment, after inputting the modeling logic information into the second large model and obtaining the modeling process information and modeling detailed operation information output by the second large model, the method further includes: Step 1020, performing a second verification on the modeling process information and the modeling detailed operation information; Step 1021, if the second verification fails, modify the modeling process information and the modeling detailed operation information based on the second verification result; Step 1022, input the modified modeling process information and the modified modeling detailed operation information into the third largest model to obtain the modeling language output by the third largest model.

[0091] It is understandable that the verification is to ensure that the modeling process information and modeling detailed operation information are reasonable and feasible in actual application. The modeling process information specifies the order of modeling steps. If there is an error in it, for example, in the modeling of cylindrical parts, the cylindrical body should be created first and then the holes should be punched, but the order in the modeling process information is reversed, the actual modeling will cause the model to fail to build or not meet the design requirements. The modeling detailed operation information involves specific operations in the CAD software. If the operation information is inaccurate, such as the punching parameters set exceed the size range of the cylinder, it will also cause problems with the model.

[0092] A series of rules based on industry standards, CAD software features, and previous modeling experience can be formulated for verification. If the verification fails, the modeling process information and modeling detailed operation information are modified based on the verification results. For example, if it is a problem with the modeling process information, the steps are rearranged in the correct order or the missing steps are supplemented; if it is a problem with the modeling detailed operation information, the parameters are reset and the operation tools are replaced according to the actual situation and relevant standards. Finally, the modified modeling process information and the modified modeling detailed operation information are input into the third model to obtain the modeling language output by the third model.

[0093] Based on the above embodiment, after inputting the modeling requirement information into the first large model and obtaining the modeling logic information output by the first large model, the method further includes: Step 1030, matching parts from a parts library according to the modeling logic information; the parts library includes modeling information of the CAD model that has been completed; Step 1031, when a target part is matched, obtaining modeling process information, modeling detailed operation information and modeling language of the target part from the parts library; Step 1032, fine-tuning the modeling process information, modeling detailed operation information and modeling language of the target part according to the difference information between the preset modeling requirements and the target part, to obtain target modeling process information, target modeling detailed operation information and target modeling language; Step 1033, sending the target modeling process information and the target modeling detailed operation information to the second large model, and sending the target modeling language to the third large model to generate the CAD model.

[0094] The parts library includes the modeling information of the completed CAD models, such as modeling logic information, modeling process information, modeling detailed operation information and modeling language, etc. It is understandable that the system is constantly carrying out new modeling work, and models of different types and structures and their related modeling supporting information will be continuously added to the parts library, so that the parts library is constantly expanded, the content is richer and more complete, and then the probability of retrieving suitable similar parts and their corresponding information when encountering similar modeling needs in the future is increased, and the overall efficiency of modeling is further improved.

[0095] According to the modeling logic information, parts are matched from the parts library; if the target part is matched, such as a part with a similarity greater than a set threshold (such as 98%); the modeling process information, modeling detailed operation information and modeling language of the target part are obtained from the parts library; then, according to the difference information between the preset modeling requirements and the target part, the modeling process information, modeling detailed operation information and modeling language of the target part are fine-tuned to obtain the target modeling process information, target modeling detailed operation information and target modeling language; finally, the target modeling process information and target modeling detailed operation information are sent to the second largest model, and the target modeling language is sent to the third largest model to generate a CAD model. For example, the second largest model and the third largest model can generate a CAD model directly through the target modeling language without processing the received information.

[0096] The embodiment of the present invention utilizes the existing part resources in the parts library and the synergy of multiple large models to efficiently generate a CAD model process, which can quickly create a model when there are suitable target parts to match, thereby improving the efficiency of the overall modeling work.

[0097] The embodiments of the present invention can ensure the accuracy of the modeling process information and the modeling detailed operation information through verification, thereby improving the modeling accuracy.

[0098] In order to further analyze and illustrate the intelligent modeling method proposed in the present invention, refer to the following examples.

[0099] The embodiment of the present invention specifically provides an intelligent modeling method based on a large model, aiming to improve design work efficiency, accelerate the innovation process, lower the entry threshold in the design field, and promote innovation and breakthroughs in the design field.

[0100] refer to Figure 2 The basic idea of ​​the intelligent modeling method based on large models is to use the language understanding ability of large models, input professional data related to design / modeling for training and fine-tuning, gradually standardize the spoken demand expression and finally form software-recognizable code. This method involves three large models, namely large model 1 (i.e. the first large model), large model 2 (i.e. the second large model) and large model 3 (i.e. the third large model). The main information of each model is as follows: Large model 1: The training and fine-tuning data is the CAD model with the modeling order and the corresponding modeling specifications. The basic rules and order of modeling are learned (for example, stretching out a disk first and then punching holes in it, rather than the other way around), forming the modeling logic, and thus converting the modeling requirements into the modeling logic.

[0101] Big Model 2: The data for training and fine-tuning is the CAD software tutorial, which simulates the process of designers learning the software and uses its language understanding ability to decompose and convert the modeling process output by Big Model 1 into the corresponding required feature combinations and detailed operations corresponding to each feature.

[0102] Big Model 3: The data for training and fine-tuning are the programming syntax of the CAD background code and the related code library. We learn the correspondence rules between modeling operations and the software background code, use the big model to implement the automatic encoding function, and convert the detailed operations output by Big Model 2 into a language (code) that can be recognized by the CAD software.

[0103] In order to better divide the work among the models and thus achieve better results, the embodiment of the present invention uses a method of converting multiple large models step by step.

[0104] For example, Figure 2 Take the parts in as an example to illustrate the overall automatic modeling process: 1. The designer expresses the following modeling requirements by voice or text: He wants a part with a flange at one end of a round tube. The length of the round tube is 100, the outer diameter is 100, the inner diameter is 80, the flange thickness is 20, the outer diameter is 180, there are 6 light holes evenly distributed on the flange, the pitch circle is 140, and the aperture is 12.

[0105] The above modeling requirements can be given all at once or in batches, and modeled step by step through human-computer interaction. The implementation logic of the two is basically similar. The subsequent steps describe the automatic modeling process in a one-time form.

[0106] 2. After receiving the above modeling requirements, the large model 1 will generate a modeling process that conforms to the normal modeling logic, as shown below: 2.1) Stretch out a round tube; 2.2) Stretch a flange at one end of the round tube; 2.3) Drill holes in the flange.

[0107] The large model 1 can provide templated statements for modeling, and can also search for parts with similar historical modeling requirements in the model library for modification on this basis, thus standardizing and unifying part modeling. At the same time, it supports contextual understanding so that designers can further propose modification requirements based on the generated results, making it easier to gradually build more complex parts.

[0108] 3. The large model 2 will be converted into the following modeling process and detailed operations according to the modeling process in the previous step. Since the specific operation process of the software is relatively cumbersome and lengthy, the following is a detailed explanation of the stretching of the round tube part as an example: 3.1) Stretched round tube, outer diameter Փ100, inner diameter Փ80, height 100: 3.11) Select "Extrude" in the feature; 3.12) Select the front view reference plane as the sketch plane; 3.13) Select the "Circle" in the sketch, select the origin as the center of the circle, and enter 50 (100 / 2) in the "Radius" item of "Parameters"; 3.14) Select the origin as the center of the circle again, enter 40 (80 / 2) in the "Radius" item of "Parameters", and then exit the sketch; 3.15) Enter 100 in the "Extrusion Depth" item and confirm.

[0109] 3.2) Stretched flange, outer diameter Փ180, height 20.

[0110] 3.3) Drill a hole on the flange with a hole diameter of Փ12, a pitch circle diameter of Փ140, and a diameter ring array of 6.

[0111] 4. Large Model 3 has learned the correspondence rules between manual modeling operations and background codes of CAD software, and then used large model technology for automatic coding to convert the above modeling operations into codes corresponding to each manual operation step (which can be codes or languages ​​that other software can recognize).

[0112] 5. Through CAD secondary development technology, develop a CAD interface to import the code from the previous step.

[0113] 6. The CAD software receives the code entered in the previous step and uses its own modeling function to complete the modeling.

[0114] 7. The designer proposes modification suggestions based on the generated model and continuously improves the model until the desired model is achieved or further modeling requirements are increased.

[0115] The embodiments of the present invention realize the automation of design and modeling by utilizing large-scale artificial intelligence models, significantly improving design efficiency and reducing learning costs, while enabling designers to focus on creative ideas, reducing the time consumption of manual modeling, and ensuring the integrity and accuracy of ideas. In addition, this technology simplifies the software operation process, lowers the entry threshold in the design field, enables more people to participate in the design process, and thus promotes the development of innovation.

[0116] The intelligent modeling device provided by the present invention is described below. The intelligent modeling device described below and the intelligent modeling method described above can be referenced to each other.

[0117] refer to Figure 3 The intelligent modeling device provided by the present invention includes a first modeling module 301, a second modeling module 302, a third modeling module 303 and a CAD model generating module 304.

[0118] The first modeling module 301 is used to input modeling requirement information into a first large model to obtain modeling logic information output by the first large model; The second modeling module 302 is used to input the modeling logic information into the second large model to obtain the modeling process information and modeling detailed operation information output by the second large model; A third modeling module 303, used for inputting the modeling process information and the modeling detailed operation information into a third large model to obtain a modeling language output by the third large model; The CAD model generation module 304 is used to generate a computer-aided design CAD model based on the modeling language; wherein the first large model is trained based on a CAD model carrying a modeling sequence and modeling specification information; the second large model is trained based on modeling software tutorial information; and the third large model is trained based on the programming syntax and code library required for modeling.

[0119] The intelligent modeling device provided by the embodiment of the present invention inputs modeling requirement information into the first large model to obtain modeling logic information output by the first large model; inputs modeling logic information into the second large model to obtain modeling process information and modeling detailed operation information output by the second large model; inputs modeling process information and modeling detailed operation information into the third large model to obtain modeling language output by the third large model; based on the modeling language, a computer-aided design CAD model is generated; wherein the first large model is obtained by training based on a CAD model carrying a modeling sequence and modeling specification information; the second large model is obtained by training based on modeling software tutorial information; and the third large model is obtained by training based on programming syntax and code library required for modeling. The present invention uses large models to replace manual modeling, realizes the automation and rapid creation of models, reduces manual modeling time, and improves the efficiency and quality of modeling design work.

[0120] In one embodiment, the second modeling module 302 is further configured to: Perform a first check on the modeling logic information; if the first check fails, modify the modeling logic information based on the first check result; input the modified modeling logic information into the second large model to obtain modeling process information and modeling detailed operation information output by the second large model.

[0121] In one embodiment, the third modeling module 303 is further used to: Perform a second verification on the modeling process information and the modeling detailed operation information; if the second verification fails, modify the modeling process information and the modeling detailed operation information based on the second verification result; input the modified modeling process information and the modified modeling detailed operation information into the third large model to obtain the modeling language output by the third large model.

[0122] In one embodiment, the third modeling module 303 is further used to: According to the modeling logic information, parts are matched from a parts library; the parts library includes modeling information of CAD models that have been completed; when a target part is matched, the modeling process information, modeling detailed operation information and modeling language of the target part are obtained from the parts library; according to the difference information between the preset modeling requirements and the target part, the modeling process information, modeling detailed operation information and modeling language of the target part are fine-tuned to obtain the target modeling process information, target modeling detailed operation information and target modeling language; the target modeling process information and the target modeling detailed operation information are sent to the second large model, and the target modeling language is sent to the third large model to generate the CAD model.

[0123] In one embodiment, the first large model includes a conversion layer; a first modeling module 301, which is used to: The modeling requirement text is input into the conversion layer for modeling logic conversion to obtain the modeling logic information output by the conversion layer.

[0124] In one embodiment, the second large model includes a decomposition layer and a conversion layer; the second modeling module 302 is used to: The modeling logic information is input into the decomposition layer to decompose the modeling process, and the modeling process information output by the decomposition layer is obtained; the modeling process information is input into the conversion layer to convert the operation information, and the modeling detailed operation information of each step of the modeling process is obtained by the conversion layer.

[0125] In one embodiment, the third large model includes a coding layer; a third modeling module 303, which is used to: The modeling process information and the modeling detailed operation information are input into the encoding layer for automatic encoding to obtain the modeling language output by the encoding layer.

[0126] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the intelligent modeling method, which includes: inputting the modeling requirement information into the first large model to obtain the modeling logic information output by the first large model; inputting the modeling logic information into the second large model to obtain the modeling process information and the modeling detailed operation information output by the second large model; inputting the modeling process information and the modeling detailed operation information into the third large model to obtain the modeling language output by the third large model; based on the modeling language, generating a computer-aided design CAD model; wherein the first large model is obtained by training based on the CAD model carrying the modeling sequence and the modeling specification information; the second large model is obtained by training based on the modeling software tutorial information; and the third large model is obtained by training based on the programming grammar and code library required for modeling.

[0127] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0128] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the intelligent modeling method provided by the above-mentioned methods, and the method includes: inputting modeling requirement information into a first large model to obtain modeling logic information output by the first large model; inputting the modeling logic information into a second large model to obtain modeling process information and modeling detailed operation information output by the second large model; inputting the modeling process information and the modeling detailed operation information into a third large model to obtain a modeling language output by the third large model; based on the modeling language, generating a computer-aided design CAD model; wherein, the first large model is trained based on a CAD model carrying a modeling sequence and modeling specification information; the second large model is trained based on modeling software tutorial information; and the third large model is trained based on the programming syntax and code library required for modeling.

[0129] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the intelligent modeling method provided by the above-mentioned methods, the method comprising: inputting modeling requirement information into a first large model to obtain modeling logic information output by the first large model; inputting the modeling logic information into a second large model to obtain modeling process information and modeling detailed operation information output by the second large model; inputting the modeling process information and the modeling detailed operation information into a third large model to obtain a modeling language output by the third large model; based on the modeling language, generating a computer-aided design CAD model; wherein, the first large model is trained based on a CAD model carrying a modeling sequence and modeling specification information; the second large model is trained based on modeling software tutorial information; and the third large model is trained based on the programming grammar and code library required for modeling.

[0130] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0131] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent modeling method, characterized in that: include: Inputting modeling requirement information into the first large model to obtain modeling logic information output by the first large model; Inputting the modeling logic information into the second largest model to obtain modeling process information and modeling detailed operation information output by the second largest model; Inputting the modeling process information and the modeling detailed operation information into a third large model to obtain a modeling language output by the third large model; Based on the modeling language, a computer-aided design (CAD) model is generated; Among them, the first large model is trained based on the CAD model carrying the modeling sequence and modeling specification information; the second large model is trained based on the modeling software tutorial information; and the third large model is trained based on the programming syntax and code library required for modeling.

2. The intelligent modeling method according to claim 1, characterized in that: After inputting the modeling requirement information into the first large model and obtaining the modeling logic information output by the first large model, the method further includes: Performing a first check on the modeling logic information; If the first verification fails, modifying the modeling logic information based on the first verification result; The modified modeling logic information is input into the second large model to obtain the modeling process information and modeling detailed operation information output by the second large model.

3. The intelligent modeling method according to claim 1, characterized in that: After inputting the modeling logic information into the second large model and obtaining the modeling process information and modeling detailed operation information output by the second large model, the method further includes: Performing a second verification on the modeling process information and the modeling detailed operation information; If the second verification fails, modifying the modeling process information and the modeling detailed operation information based on the second verification result; The modified modeling process information and the modified modeling detailed operation information are input into the third largest model to obtain the modeling language output by the third largest model.

4. The intelligent modeling method according to claim 1, characterized in that: After inputting the modeling requirement information into the first large model and obtaining the modeling logic information output by the first large model, the method further includes: Matching parts from a parts library according to the modeling logic information; the parts library includes modeling information of the CAD models that have been completed; When the target part is matched, the modeling process information, detailed modeling operation information and modeling language of the target part are obtained from the parts library; According to the difference information between the preset modeling requirements and the target part, fine-tune the modeling process information, the modeling detailed operation information and the modeling language of the target part to obtain the target modeling process information, the target modeling detailed operation information and the target modeling language; The target modeling process information and the target modeling detailed operation information are sent to the second large model, and the target modeling language is sent to the third large model to generate the CAD model.

5. The intelligent modeling method according to claim 1, characterized in that: The first large model includes a conversion layer; the step of inputting the modeling requirement information into the first large model to obtain the modeling logic information output by the first large model includes: The modeling requirement text is input into the conversion layer for modeling logic conversion to obtain the modeling logic information output by the conversion layer.

6. The intelligent modeling method according to claim 1, characterized in that: The second largest model includes a decomposition layer and a transformation layer; The step of inputting the modeling logic information into the second large model to obtain the modeling process information and modeling detailed operation information output by the second large model includes: Inputting the modeling logic information into the decomposition layer to decompose the modeling process, and obtaining the modeling process information output by the decomposition layer; The modeling process information is input into the conversion layer for operation information conversion, and the modeling detailed operation information of each step of the modeling process output by the conversion layer is obtained.

7. The intelligent modeling method according to claim 1, characterized in that: The third model includes a coding layer; the inputting the modeling process information and the modeling detailed operation information into the third model to obtain the modeling language output by the third model includes: The modeling process information and the modeling detailed operation information are input into the encoding layer for automatic encoding to obtain the modeling language output by the encoding layer.

8. An intelligent modeling device, characterized in that: include: A first modeling module, used for inputting modeling requirement information into a first large model to obtain modeling logic information output by the first large model; A second modeling module, used for inputting the modeling logic information into a second large model, and obtaining modeling process information and modeling detailed operation information output by the second large model; A third modeling module, used for inputting the modeling process information and the modeling detailed operation information into a third large model to obtain a modeling language output by the third large model; The CAD model generation module is used to generate a computer-aided design CAD model based on the modeling language; wherein the first large model is trained based on a CAD model carrying a modeling sequence and modeling specification information; the second large model is trained based on modeling software tutorial information; and the third large model is trained based on the programming syntax and code library required for modeling.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the intelligent modeling method as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the intelligent modeling method as claimed in any one of claims 1 to 7 is implemented.

Images