A three-dimensional design real-time verification method and system based on artificial intelligence

Abstract

The application discloses a kind of based on artificial intelligence three-dimensional design real-time verification method and system, it is related to computer aided design technical field.The method is configured event listening interface in three-dimensional design environment, real-time monitoring user's editing instruction to assembly body;Parse instruction and extract the geometric features of target component and the attribute metadata containing historical engineering change frequency;The extracted data are input into the pre-trained neural network model to carry out risk inference;When risk score exceeds threshold value, obtain component coordinates and superimposed rendering visual warning mark in graphical user interface.The application drives AI model using historical data, preposes verification link, realizes the instant prediction and intelligent repair recommendation of design risk, solves the problem of traditional simulation feedback lag, significantly reduces the cost of research and development trial and error.

Description

Technical Field

[0001] This invention relates to the fields of computer-aided design (CAD) and computer-aided engineering (CAE), and in particular to a method and system for real-time risk assessment and verification of three-dimensional assembly design using artificial intelligence technology. Background Technology

[0002] In the product development process of modern manufacturing, 3D modeling (design) and simulation verification (analysis) are usually two independent and separate stages. After completing the preliminary design in CAD software, the designer needs to export the model and hand it over to the CAE engineer for finite element analysis (FEA) or design for manufacturability (DFM) checks.

[0003] This traditional linear process has significant drawbacks:

[0004] First, there is a lag in feedback. Simulation analysis typically takes hours or even days, which means that design flaws (such as interference, stress concentration, and inappropriate material selection) cannot be detected in a timely manner.

[0005] Second, data utilization is low. Enterprises have accumulated a wealth of valuable engineering data on "rework," "changes," and "failures" in their historical projects, but existing CAD software only focuses on geometric modeling and cannot use this historical data to predict the risks of current designs.

[0006] Third, the cost of trial and error is high. According to the "1-10-100 rule", errors that are not discovered in the design stage will have exponentially higher correction costs once they are transferred to the manufacturing stage.

[0007] In existing technologies, although some software attempts to integrate simple rule checks, they are often based on fixed geometric parameters, lack the ability to learn from historical experience data, and are difficult to provide millisecond-level real-time feedback while the user is operating. Summary of the Invention

[0008] This invention aims to solve the aforementioned technical problems by providing an artificial intelligence-based real-time verification method and system for 3D design. This method, through real-time monitoring of design operations and combined with an AI model driven by historical engineering data, achieves immediate prediction and visual warnings of design risks, thereby bringing the verification process forward and reducing R&D trial-and-error costs.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] A real-time verification method for 3D design based on artificial intelligence includes the following steps:

[0011] Configure an event listening interface in the 3D computer-aided design environment to monitor the editing operation commands issued by the user for the 3D assembly in real time;

[0012] in response to detecting the editing operation instruction, parsing the instruction to lock the affected target component, and extracting geometric feature data and attribute metadata of the target component; the attribute metadata at least includes a historical engineering change frequency parameter and a manufacturing complexity parameter of the target component;

[0013] inputting the extracted geometric feature data and attribute metadata into a pre-trained risk assessment model for inference analysis to obtain a risk score of the target component; the risk assessment model is a neural network model trained based on historical engineering project data;

[0014] comparing the risk score with a preset risk threshold to generate a health degree assessment result, if the health degree assessment result indicates that there is a risk, obtaining the three-dimensional spatial coordinates of the target component, and generating a corresponding visual warning mark, and superimposing and rendering the visual warning mark to a graphical user interface of the three-dimensional computer-aided design environment.

[0015] Further, the attribute metadata further includes at least one of the following data: supply chain inventory state data, raw material compatibility data, tolerance level data, and historical rework record data;

[0016] The extracted geometric feature data of the target component specifically includes: calculating the three-dimensional bounding box volume, surface area, topological surface number, and interference volume with other components of the target component.

[0017] Further, the training process of the risk assessment model includes:

[0018] retrieve historical component data and associated engineering change order records from a product lifecycle management system;

[0019] perform feature cleaning and labeling on the historical component data, mark components associated with engineering change orders due to design defects as risk samples, and mark components not associated with the engineering change orders as safe samples;

[0020] construct a neural network model, use the risk samples and safe samples to supervise the training of the neural network model, minimize the loss function between the predicted results and the actual change records, and obtain the risk assessment model.

[0021] Further, the superimposed rendering of the visual warning mark to the graphical user interface of the three-dimensional computer-aided design environment specifically includes:

[0022] based on the three-dimensional geometric data of the target component, calculating the vertex coordinates of the minimum circumscribed bounding box;

[0023] creating a transparent layer independent of the original model data in a rendering pipeline of the three-dimensional computer-aided design environment;

[0024] drawing a three-dimensional mesh body with a preset color and transparency on the transparent layer according to the vertex coordinates to wrap and display the target component.

[0025] Further, the method further comprises:

[0026] when the health assessment result indicates that there is a risk, searching in a component library based on the attribute metadata to filter out alternative component information with a lower historical engineering change frequency;

[0027] generating an intelligent repair suggestion containing the alternative component information in the graphical user interface;

[0028] in response to receiving an instruction of the user confirming to adopt the intelligent repair suggestion, automatically replacing the target component in the three-dimensional assembly with the alternative component.

[0029] Further, the event listening interface is specifically configured to listen to node addition events, node removal events, node transformation matrix update events, and node attribute change events in a scene graph.

[0030] The real-time monitoring step further comprises: setting a jitter prevention time window, and only when no new editing operation instruction is detected within the time window, triggering the subsequent data extraction step.

[0031] Further, the health assessment result comprises a quantitative risk probability value and a qualitative risk description text.

[0032] The visual warning mark further comprises a text label displayed around the target component, and the text label is used to display the risk description text.

[0033] In addition, the present application also provides a three-dimensional design real-time verification system based on artificial intelligence, comprising:

[0034] a monitoring module configured to configure an event listening interface in a three-dimensional computer-aided design environment, and monitor editing operation instructions issued by a user for a three-dimensional assembly in real time;

[0035] a data processing module configured to, in response to detecting the editing operation instruction, parse the instruction to lock a target component affected, and extract geometric feature data and attribute metadata of the target component;

[0036] The inference analysis module is configured to input the extracted geometric feature data and attribute metadata into a pre-trained risk assessment model for inference analysis, so as to obtain a risk score of the target component.

[0037] The rendering interaction module is configured to compare the risk score with a preset risk threshold to generate a health degree evaluation result, and to superimpose and render a visual warning mark at the three-dimensional spatial coordinates of the target component in the graphical user interface of the three-dimensional computer-aided design environment according to the result.

[0038] The present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor.

[0039] The present application also provides a computer-readable storage medium having a computer program stored thereon.

[0040] The present application has the following advantages:

[0041] 1. Decision preposition: The verification process is advanced from after the design is completed to when the design is being performed, helping designers to know potential manufacturing and supply chain risks in an instant of “drawing”.

[0042] 2. Data-driven: Unlike traditional physical simulation, the present application uses historical rework rate metadata for inference, which can discover “empirical” risks that cannot be identified by physical simulation.

[0043] 3. Efficient and intuitive: Through three-dimensional spatial superposition rendering, designers can intuitively locate problem areas without switching software windows, greatly improving collaborative design efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 is a method flowchart provided by the first embodiment of the present application;

[0045] Figure 2 is a risk assessment model training and inference logic diagram provided by the first embodiment of the present application;

[0046] Figure 3 is a system architecture block diagram provided by the third embodiment of the present application;

[0047] Figure 4 is a schematic diagram of risk visualization display in a graphical user interface provided by the first embodiment of the present application. DETAILED DESCRIPTION

[0048] The present application will be further described in detail below in combination with the drawings and specific embodiments.

[0049] Embodiment One

[0050] AsFigure 1 As shown, the embodiment provides a three-dimensional design real-time verification method based on artificial intelligence, applied to a computer terminal installed with three-dimensional CAD software (such as SolidWorks, Catia or Web-based 3D design platform). The method specifically includes the following steps:

[0051] Step S1: Establish event listening mechanism

[0052] The system runs a daemon or plug-in in the background of the CAD software, which is mounted on the event bus of the scene graph (Scene Graph) through the API interface. The system is configured to listen to the following core events: NodeAdded (node addition), NodeRemoved (node removal), NodeTransformUpdated (position / rotation change).

[0053] In order to avoid performance loss caused by frequent calculation, the system sets a debounce (Debounce) time window, for example, 200 milliseconds. That is, only when the user stops operating for more than 200 milliseconds, the subsequent verification process is triggered.

[0054] Step S2: Extract feature data

[0055] When it is monitored that a user drags a new component (for example, "transmission gear A") into the assembly, the system automatically locks the target component and extracts two types of data:

[0056] 1. Geometric feature data: including the length, width and height dimensions, volume, surface area and number of faces (Face) of the minimum circumscribed bounding box (AABB) of the component.

[0057] 2. Attribute metadata: the system initiates an asynchronous query to the enterprise's PLM (Product Lifecycle Management) database through the unique identifier (Part ID) of the component to obtain the non-geometric attributes of the component.

[0058] 3. In this embodiment, the extracted attribute metadata is formatted as the following JSON structure for algorithm call:

[0059] {

[0060] "component_id": "GEAR-2024-X",

[0061] "material_code": "STEEL_45",

[0062] "history_stats": {

[0063] "total_usage_count": 500, / / historical usage times

[0064] "eco_change_count": 75, / / Number of historical engineering change

[0065] "rework_rate": 0.15 / / Historical rework rate (15%)

[0066] },

[0067] "supply_chain": {

[0068] "stock_level": "LOW", / / Stock level

[0069] "lead_time_days": 45 / / Lead time in days

[0070] },

[0071] "complexity_score": 0.85 / / Manufacturing complexity score

[0072] }

[0073] Step S3: AI model inference

[0074] As shown in FIG. 2, a risk assessment model training and inference logic diagram provided by an embodiment of the present application is shown. The data extracted above is vectorized and input into a pre-trained "risk assessment model".

[0075] Model construction and training: The model adopts a multi-layer perceptron (MLP) combined with a random forest (Random Forest) architecture.

[0076] 1. Input layer: receives normalized geometric parameters (volume, complexity) and metadata parameters (rework rate, inventory status).

[0077] 2. Output layer: outputs a confidence score (Risk Score) between 0 and 1, and a classification label of the risk type (such as "high rework risk", "out-of-stock risk", "processing difficulty").

[0078] 3. Training data source: the training set comes from the enterprise's historical project database in the past 5 years. The feature value X is the component attribute, and the label value Y is whether the component has triggered ECO (Engineering Change Order) in the subsequent process. Through supervised learning, the model learns the implicit rules of "what kind of components are prone to errors in what combination".

[0079] In this embodiment, assume the model outputs for "Gear A" a Risk Score = 0.82 and a risk type of High_Rework_Rate.

[0080] Step S4: Visual rendering feedback

[0081] The system determines that the Risk Score is greater than a preset threshold (e.g., 0.7), triggering an alert process.

[0082] 1. Coordinate acquisition: Obtain the bounding box vertex coordinates of "Gear A" in the current window coordinate system.

[0083] 2. Layer superposition: In a three-dimensional rendering engine (such as Three.js or OpenGL), without modifying the original model data, create a temporary transparent layer.

[0084] 3. Marker drawing: In the position corresponding to this layer, draw a red, semi-transparent (Alpha = 0.4) cube mesh that wraps around the gear. At the same time, render a floating HTML information card next to the gear through a screen space projection algorithm, displaying the text: "Warning: the historical rework rate of this component is as high as 15%, please check the tolerance fit."

[0085] 4. As shown in Figure 4 , the designer can see the warning box at the moment of completing the drag operation, immediately realizing the risk.

[0086] Embodiment Two

[0087] Based on the method of Embodiment One, this embodiment also provides a Smart Fix function.

[0088] When the system detects that "Gear A" has a high risk, the AI model will further perform vector similarity retrieval in the component library to find alternative components that have similar geometric shapes but "better historical performance" (i.e., low rework rate, sufficient inventory).

[0089] For example, the system retrieves "Gear B", which has a geometric parameter similarity of 99% to A, but a rework rate of only 1%.

[0090] The system displays "Recommended replacement: Gear B" in the alert card. After the user clicks the "One-click replacement" button, the system automatically calls the CAD interface, unloads component A, loads component B, and maintains the original assembly constraint relationship unchanged.

[0091] Embodiment Three

[0092] As shown in Figure 3As shown, the application also provides an artificial intelligence-based three-dimensional design real-time verification system, comprising:

[0093] 1. Monitoring module: resides in the design software of the user terminal, responsible for monitoring mouse and keyboard events, and capturing design intent.

[0094] 2. Data processing module: responsible for connecting enterprise databases (PLM / ERP), fusing unstructured geometric data and structured business data, and generating feature vectors.

[0095] 3. Inference analysis module: deployed on a cloud server or a local high-performance workstation, running a trained neural network model to provide millisecond-level inference services.

[0096] 4. Rendering interaction module: responsible for converting abstract risk data into intuitive three-dimensional graphics and UI elements to achieve an augmented reality (AR)-like interactive experience.

[0097] Those skilled in the art can understand that the "historical rework rate" used in the above embodiments is only a typical risk indicator. In actual application, multi-dimensional data such as "raw material price fluctuation trend" and "supplier credit rating" can also be combined for comprehensive evaluation, which all belong to the protection scope of the present application.

[0098] The above is only a preferred embodiment of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A real-time verification method for 3D design based on artificial intelligence, characterized in that, Includes the following steps: Configure an event listening interface in the 3D computer-aided design environment to monitor the editing operation commands issued by the user for the 3D assembly in real time; In response to detecting the edit operation instruction, the instruction is parsed to locate the affected target component, and the geometric feature data and attribute metadata of the target component are extracted; the attribute metadata includes at least the historical engineering change frequency parameter and manufacturing complexity parameter of the target component; The extracted geometric feature data and attribute metadata are input into a pre-trained risk assessment model for inference analysis to obtain the risk score of the target component; The risk assessment model is a neural network model trained based on historical engineering project data. The risk score is compared with a preset risk threshold to generate a health assessment result. If the health assessment result indicates that there is a risk, the three-dimensional spatial coordinates of the target component are obtained, and a corresponding visual warning mark is generated. The visual warning mark is overlaid and rendered onto the graphical user interface of the three-dimensional computer-aided design environment.

2. The method according to claim 1, characterized in that, The attribute metadata also includes at least one of the following data: supply chain inventory status data, raw material compatibility data, tolerance level data, and historical rework record data. The extraction of geometric feature data of the target component specifically includes: calculating the three-dimensional bounding box volume, surface area, number of topological surfaces, and interference volume with other components of the target component.

3. The method according to claim 1, characterized in that, The training process of the risk assessment model includes: Retrieve historical component data and associated engineering change order records from the product lifecycle management system; The historical component data is cleaned and labeled with features. Components associated with engineering change orders due to design defects are marked as risk samples, and components not associated with engineering change orders are marked as safe samples. A neural network model is constructed, and the model is trained under supervision using the risk samples and safety samples. The risk assessment model is obtained by minimizing the loss function between the predicted results and the actual change records.

4. The method according to claim 1, characterized in that, The step of overlaying and rendering the visual warning marker onto the graphical user interface of the 3D computer-aided design environment specifically includes: Based on the three-dimensional geometric data of the target component, calculate the vertex coordinates of its minimum bounding box; Create a transparent layer in the rendering pipeline of the 3D computer-aided design environment that is independent of the original model data; On the transparent layer, a three-dimensional mesh with preset color and transparency is drawn according to the vertex coordinates to wrap and display the target component.

5. The method according to claim 1, characterized in that, The method further includes: When the health assessment result indicates a risk, the component library is searched based on the attribute metadata to filter out alternative component information with a lower frequency of historical engineering changes. The graphical user interface generates intelligent repair suggestions containing information about the alternative components. In response to receiving a user's confirmation of acceptance of the smart repair suggestion, the target component in the 3D assembly is automatically replaced with the alternative component.

6. The method according to claim 1, characterized in that, The event listening interface is specifically configured to listen for node addition events, node removal events, node transformation matrix update events, and node attribute change events in the scene graph. The real-time monitoring step also includes: setting a stabilization time window, and triggering the subsequent data extraction step only when no new editing operation command is detected within the time window.

7. The method according to claim 1, characterized in that, The health assessment results include quantitative risk probability values ​​and qualitative risk description text; The visual warning label also includes a text label that floats around the target component, the text label being used to display the risk description text.

8. A real-time verification system for 3D design based on artificial intelligence, characterized in that, include: The monitoring module is configured to set up an event listening interface in the 3D computer-aided design environment to monitor the editing operation commands issued by the user for the 3D assembly in real time. The data processing module is configured to, in response to the detection of the editing operation instruction, parse the instruction to lock the affected target component, and extract the geometric feature data and attribute metadata of the target component; The reasoning and analysis module is configured to input the extracted geometric feature data and attribute metadata into a pre-trained risk assessment model for reasoning and analysis to obtain the risk score of the target component. The rendering interaction module is configured to compare the risk score with a preset risk threshold to generate a health assessment result, and based on the result, to overlay a rendered visual warning mark at the three-dimensional spatial coordinates of the target component in the graphical user interface of the three-dimensional computer-aided design environment.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 7.

Images