Interactive digital twinning enhancement modeling method

Abstract

The invention discloses an interactive digital twinning enhancement modeling method. The method comprises the following steps: constructing a digital twinning model which is dynamically updated in real time; building a natural interaction interface system comprising a voice interaction module, a gesture recognition module and a tactile feedback module; a simulation calculation result is fed back in a three-dimensional visualization mode, meanwhile, virtual physical attributes of the operation object are output through a tactile feedback module, and optimized parameters are synchronized to a physical entity control system according to a user confirmation instruction. The method has the beneficial effects that according to the scheme, through data layer optimization, natural interaction, precise simulation, visual feedback and self-optimization closed loop, the interactivity, accuracy and practicability of the digital twin model are comprehensively improved, and powerful technical support is provided for optimization decision making in a complex scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital twinning, in particular to an interactive digital twinning enhanced modeling method. BACKGROUND

[0002] As a core technology for realizing real-time mapping and dynamic interaction between the physical world and the virtual space, digital twinning technology has been widely applied in key fields such as industrial production, equipment operation and maintenance, city management, and medical health. Through accurate modeling of the geometric shape, physical properties, behavior rules, and running state of physical entities, digital twinning models can provide visual and quantifiable analysis tools for process optimization, fault prediction, and decision support, greatly improving the management efficiency and intelligent level of complex systems.

[0003] However, the existing digital twinning modeling methods still have the following significant deficiencies in practical application: The running data of physical entities usually comes from multiple channels such as real-time collection by sensors, historical business systems, three-dimensional model files, and operation logs. The data types are complex (structured, unstructured, time series data, etc.) and have problems such as noise, missing, and redundancy. Traditional modeling methods lack efficient multi-source data association storage mechanisms, have low data fusion efficiency, and have insufficient ability to repair low-quality data, which limits the accuracy and reliability of the twinning model and makes it difficult to truly reflect the state of the physical entity.

[0004] The interaction of existing digital twinning systems mainly relies on traditional input devices such as mice, keyboards, or touch screens, which have a cumbersome operation process and lack natural interaction capabilities. Users need to master professional software operation skills and cannot easily make real-time adjustments to the model through intuitive methods such as voice and gestures. In complex scenarios such as industrial production, management personnel cannot quickly respond to dynamically changing production demands, resulting in low interaction efficiency.

[0005] Simulation calculation is a core link for digital twinning models to support decision-making, but existing methods have the contradiction between simulation accuracy and response speed: fixed time step in simple operation scenarios leads to resource waste, and simulation takes too long in complex process reconstruction. At the same time, feedback methods mainly rely on two-dimensional charts or static three-dimensional displays, lacking immersive feedback on physical properties (such as hardness and resistance) of the operating object, making it difficult for users to intuitively perceive the adjustment effect and affecting decision-making efficiency.

[0006] The performance improvement of digital twinning models highly depends on human experience and lacks systematic learning of user operation behavior and feedback data from physical entities. When the application scenario or demand changes, the model cannot autonomously recommend the optimal adjustment scheme, leading to a decline in optimization efficiency over the long term and making it difficult to adapt to dynamically complex actual working conditions.

[0007] Therefore, in view of the problems of low data integration efficiency, unnatural interaction mode, non-intuitive simulation feedback and insufficient self-optimization capability in the prior art, it is necessary to develop an interactive digital twin enhanced modeling method capable of realizing efficient fusion of multi-source data, natural interaction, precise simulation feedback and continuous self-optimization, so as to improve the practicality and intelligent level of the digital twin technology and meet the fine management and control requirements of complex scenes such as industrial production. SUMMARY

[0008] The technical problem solved by the present application is to provide an interactive digital twin enhanced modeling method, which solves one or more of the above-mentioned problems in the prior art.

[0009] To solve the above technical problems, one technical solution adopted by the present application is an interactive digital twin enhanced modeling method, which is characterized by comprising the following steps:

[0010] Step 1: Construct a real-time dynamically updated digital twin model, integrate multi-source heterogeneous data through a multi-source data interface to form a multi-dimensional twin body data layer, and synchronize data between the digital twin model and the physical entity through an edge computing node, with the synchronization frequency being dynamically adjusted according to the state change rate of the entity;

[0011] Step 2: Build a natural interaction interface system including a voice interaction module, a gesture recognition module and a tactile feedback module;

[0012] Step 3: Receive the operation instruction input by the user through the natural interaction interface, and convert the operation instruction into a parameter adjustment instruction executable by the twin model;

[0013] Step 4: Perform simulation calculation on the parameter adjustment instruction based on a finite state machine model, and output the adjusted resource load distribution, production rhythm change, energy consumption fluctuation trend and potential conflict warning;

[0014] Step 5: Feedback the simulation calculation results in a three-dimensional visual manner, output the virtual physical properties of the operation object through the tactile feedback module, and synchronize the optimized parameters to the physical entity control system according to the user confirmation instruction.

[0015] In some embodiments, the construction of the multi-dimensional twin body data layer in step 1 comprises:

[0016] The multi-source data is associatedly stored by adopting a space-time index mechanism, and the space-time index takes the physical entity ID as the primary key, and associates the time stamp, spatial coordinates and attribute data;

[0017] A data quality evaluation model is established to score the integrated data in terms of completeness, accuracy and timeliness, and a filling algorithm based on a generative adversarial network (GAN) is used to repair the low-quality data, and the repair formula is:

[0018]

[0019] wherein G is a generator network, z is a random noise vector, x obs is observed incomplete data, is the repaired data.

[0020] In some embodiments, the implementation of the voice interaction module in step 2 includes:

[0021] An industry term dictionary is constructed based on a bidirectional long short-term memory network (Bi-LSTM), and the input voice signal is denoised, tokenized, and intent-recognized, with an accuracy rate of no less than 92%;

[0022] A context understanding model is established, and the current instruction and the historical dialogue sequence are associated through an attention mechanism to support cross-round instruction completion, and when a key parameter is missing, a follow-up question is automatically generated.

[0023] In some embodiments, the gesture recognition module in step 2 captures the three-dimensional coordinates and motion trajectories of the hand through a depth camera, and realizes the grabbing, moving, and scaling operations of the virtual object in the twin model.

[0024] In some embodiments, the operation instruction conversion in step 3 includes:

[0025] An operation instruction mapping library is constructed to convert natural language instructions and gesture trajectories into standardized control instructions in JSON format, including operation object ID, target parameter value, execution priority, and time constraints;

[0026] Ambiguity resolution is performed on the fuzzy instruction by comparing the current state of the twin model with the historical operation records to determine the user's true intention.

[0027] In some embodiments, the simulation calculation in step 4 includes:

[0028] A discrete event simulation engine is used to simulate the production process reconstruction instruction, calculate the waiting time, equipment utilization rate, and bottleneck process position of each process, and output the production efficiency optimization amplitude;

[0029] Based on the digital thread technology, the product life cycle data is tracked, and when the device parameters are adjusted, the influence transmission chain on the subsequent processing processes is calculated synchronously, and an influence range heat map is generated;

[0030] The time step of the simulation calculation is dynamically adjusted according to the operation complexity, with 1 second / step for simple operations and 0.1 second / step for complex process reconstruction.

[0031] In some embodiments, the three-dimensional visualization feedback in step 5 includes:

[0032] The twin model is rendered by using a level of detail (LOD) technology, when a user focuses on a specific area, the model precision of the area is automatically improved to a millimeter level, and the non-focused area is simplified to a low polygon model;

[0033] The key indicators before and after adjustment are compared and visualized, the production efficiency change is displayed by using a dynamic column chart, the resource load difference is marked by using a color gradient, and the material flow path change is simulated by using a particle special effect.

[0034] In some embodiments, the interactive digital twin enhanced modeling method further comprises a twin model self-optimization step:

[0035] Record the user interaction operation and the corresponding physical entity feedback data, and construct an operation-effect association database;

[0036] A deep reinforcement learning algorithm is used to train a recommendation model, when a user repeatedly performs a similar operation scenario, the historical optimal adjustment scheme is automatically recommended, and the reward function of the recommendation model is:

[0037] R=α·Δη+β·(1-Δ∈)-γ·t

[0038] Wherein, Δη is the production efficiency improvement rate, Δ∈ is the energy consumption growth rate, t is the operation time, and α, β and γ are weight coefficients.

[0039] In some embodiments, the natural interaction interface system further comprises an eye tracking module, the eye tracking module captures the user's gaze landing point through an infrared camera, when the gaze stays on a certain virtual object for more than a preset time (0.5-2 seconds configurable), the object is automatically highlighted and its key parameters are displayed.

[0040] In some embodiments, the parameter adjustment instruction in step 3 includes resource configuration change, production process reconstruction, and device parameter modification.

[0041] The beneficial effects of the present application are: the technical solution adopts a space-time index mechanism to realize the associated storage of multi-source heterogeneous data, combines a GAN algorithm to repair low-quality data, and ensures the integrity and accuracy of the twin body data layer. Dynamically adjust the data synchronization frequency of the edge computing node, reduce resource consumption while ensuring real-time, and solve the problems of low data integration efficiency and poor quality in traditional modeling.

[0042] The technical solution integrates voice, gesture, tactile feedback and eye tracking modules to form a multi-modal interaction system: the voice module supports industry terminology semantic understanding and cross-turn instruction completion, gesture recognition enables intuitive control of virtual objects, tactile feedback provides immersive operation experience, and eye tracking automatically highlights the object of interest and displays key parameters. Compared with traditional single interaction mode, users can efficiently adjust the model through natural actions, the intention recognition accuracy is above 92%, and the operation efficiency is significantly improved.

[0043] The technical solution converts natural language / gesture into standardized JSON instructions through an operation instruction mapping library, and combines an ambiguity resolution mechanism to clarify user intent; simulation calculation uses a discrete event engine and digital thread technology to dynamically adjust the time step (0.1-1 second / step), which can output resource load, production rhythm, energy consumption trend and conflict warning in real time, and generate an impact range heat map to provide quantitative basis for decision-making.

[0044] The technical solution realizes hierarchical rendering of three-dimensional models based on LOD technology, focusing on millimeter-level accuracy in the focus area and simplifying non-focus areas to balance performance; key indicators such as production efficiency and resource load are compared and displayed before and after adjustment through dynamic column charts, color gradients and particle special effects, and the optimization effect is intuitively presented; the parameter synchronization mechanism supports one-key delivery to physical entity control systems after user confirmation, ensuring consistency between virtual adjustment and physical execution.

[0045] The technical solution records user operation and physical feedback data to build an operation-effect association database, and uses deep reinforcement learning to train a recommendation model. When users face similar scenarios, the system automatically recommends the best historical solution (reward function considers efficiency improvement, energy consumption reduction and operation time consumption), realizes self-iteration and optimization of the twin model, and significantly reduces decision-making costs in the long term.

[0046] The technical solution can adapt to the needs of industrial production, urban management, equipment maintenance and other fields through modular design. At the same time, it supports function module extension, reserves interfaces for subsequent technology upgrade, and has good compatibility and scalability.

[0047] In summary, the scheme improves the interactivity, accuracy and practicality of digital twin models through data layer optimization, natural interaction, accurate simulation, visual feedback and self-optimization closed loop, and provides strong technical support for optimization decision-making in complex scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0049] Figure 1 The figure is a schematic diagram of the overall architecture of the present application.

[0050] Figure 2 The figure is a schematic diagram of the natural interface system of the present application.

[0051] Figure 3 The figure is a schematic diagram of receiving and converting operation instructions of the present application.

[0052] Figure 4 The figure is a schematic diagram of the self-optimization of the twin model of the present application.

[0053] Figure 5 The figure is a schematic diagram of the digital twin modeling of the automobile production workshop provided by an embodiment of the present application. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be described clearly and completely as follows. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort belong to the scope of protection of the present application.

[0055] As shown in the figure, the present embodiment takes the digital twin modeling process of a certain automobile production workshop as an example to illustrate the implementation steps of the interactive digital twin enhanced modeling method in detail. Figures 1 to 5

[0056] 1. Constructing a real-time dynamically updated digital twin model

[0057] Integrate the following multi-source heterogeneous data in the workshop through multi-source data interfaces to form a multi-dimensional twin body data layer:

[0058] Sensor real-time data stream: including welding robot temperature (±0.5℃ precision), conveyor belt speed (0-5m / s), stamping equipment pressure (0-5000kN), etc.

[0059] Historical business database: production order data in the past 3 years (such as vehicle type, output, and qualification rate), equipment maintenance records (fault type, maintenance time);

[0060] Three-dimensional scene model: 1:1 three-dimensional grid model of the workshop (including 200 devices, 5 production lines, and 3 material storage areas).​

[0061] Equipment operation log: such as the running time of servo motor, the wear degree of mechanical arm joint, etc.

[0062] Data processing method: adopt space-time index mechanism, take equipment ID (such as "welding_robot_001") as primary key, associate time stamp (accurate to millisecond), space coordinate (based on workshop coordinate system) and attribute data (such as temperature, pressure); establish data quality evaluation model, give 0-100 score to integrated data (integrity ≥ 85 points, accuracy ≥ 90 points, timeliness ≤ 5 seconds are qualified); for low quality data (such as 10% data missing caused by sensor failure), adopt GAN filling algorithm to repair:

[0063]

[0064] Among them, the generator network G is trained based on 500,000 historical complete data, the random noise vector z is 128 dimensions, and the error of the repaired data is ≤3%.

[0065] Data synchronization strategy: realize data synchronization with physical entity through edge computing node (delay ≤10ms) deployed in workshop; when the equipment is in high-speed production state (such as continuous operation of welding robot), the synchronization frequency is set to 1 time / second; when the equipment is idle, the synchronization frequency is reduced to 1 time / 10 seconds.

[0066] 2. Build natural interaction interface system

[0067] Voice interaction module: based on Bi-LSTM network, build automobile production industry term dictionary (including 3000+ professional terms, such as "AGV scheduling" "tooling fixture"), carry out noise reduction (adopt wavelet threshold denoising algorithm), word segmentation (based on jieba word segmentation tool) and intent recognition on input voice signal. The actual measurement intent recognition accuracy is 93% (test sample 1000); the context understanding model associates historical dialogue through attention mechanism, for example, the user first asks "current capacity of stamping line", and the subsequent instruction "increase by 10%" can be automatically completed as "increase the capacity of stamping line by 10%". If the instruction lacks parameters (such as "adjust robot"), the system automatically asks: "please specify the adjustment of welding robot / handling robot, and the specific parameters (speed / angle)".

[0068] Gesture recognition module: capture user's hand three-dimensional coordinates (x, y, z accuracy ±1mm) and motion trajectory through Intel RealSense D455 depth camera (frame rate 30fps, depth error ≤2%); support the following operations:

[0069] Grabbing gesture (thumb and index finger pinch): select virtual equipment model;

[0070] Translation gesture (hand uniform motion): move virtual object position (response delay ≤ 200ms);

[0071] Scaling gesture (two-finger distance change): adjust model scale (scaling range 0.1-10 times).

[0072] Haptic feedback module: use Geomagic Touch X force feedback device, when the user virtually operates the welding robot arm, output corresponding virtual physical properties:

[0073] Hardness: simulate the touch of different material workpieces (such as steel plate hardness feedback force 5-10N, rubber 2-3N);

[0074] Resistance: dynamically changes with the speed of the mechanical arm (speed 0-1m / s corresponds to resistance 1-5N).

[0075] Eye tracking module: integrate Tobii Pro Fusion eye tracker, when the user's gaze stays on the virtual stamping device for 1 second (configurable 0.5-2 seconds), automatically highlight the device model (red border + 50% transparency overlay), and display key parameters: current temperature (±1℃), cumulative stamping times (accurate to the unit place), next maintenance countdown (days: hours: minutes).

[0076] 3. Operation instruction conversion

[0077] User issues instructions through voice: "Increase production line 01 speed by 10%", system performs the following conversion:

[0078] Standardized control instruction generation: based on operation instruction mapping library, converted to JSON format: content as follows:

[0079] "Operation object ID":"production_line_01",

[0080] "Target parameter value":"speed increase 10%",

[0081] "Execution priority":"high",

[0082] "Time constraint":"execute immediately"

[0083] }

[0084] Ambiguity resolution: if the user instruction is "optimize production process", the system compares the current process (10 procedures, bottleneck in welding process) with historical operation records (similar scenarios have been optimized by "reducing 2 detection procedures"), determines that the user's intention is "reduce procedures", and generates candidate solutions for selection.

[0085] 4. Simulation calculation

[0086] Simulation of the "Line 01 speed increase 10%" instruction based on the finite state machine model:

[0087] Discrete event simulation: using the AnyLogic simulation engine, simulating the adjusted waiting time of each process (average reduction of 15 seconds), device utilization (welding robot from 75% to 82%), and bottleneck process (from welding to assembly), outputting production efficiency optimization range 8%;

[0088] Impact on the calculation of the transmission chain: based on the digital thread technology to track the whole life cycle data of the parts, after the speed adjustment, the material supply delay risk of the subsequent assembly process increases by 30%, and the red heat map marks the affected area;

[0089] Time step setting: this operation is of medium complexity, the time step is set to 0.5 seconds / step, and the simulation total time is 120 seconds (covering 2 production beats).

[0090] 5. Feedback and parameter synchronization

[0091] Three-dimensional visual feedback:

[0092] Using the Unity 3D engine to achieve Level of Detail (LOD) rendering: when the user focuses on the welding area, the model precision is improved to 0.1mm (polygon number 1 million+), and the non-focused area is simplified to 100,000 polygons;

[0093] Key indicator comparison: dynamic column chart real-time display of production efficiency before and after adjustment (from 60 units / hour to 65 units / hour), color gradient marking device load (green ≤60%, yellow 60%-80%, red >80%), particle special effect simulation of material flow path change (particle density increases by 10% after speed increase).

[0094] Haptic feedback: when the user confirms the adjustment scheme through the virtual control panel, the button feedback force changes with the button "material" (metal button 8N, plastic button 3N).

[0095] Parameter synchronization: after the user confirms, the system synchronizes the optimized parameters (production line speed from 1.2m / s to 1.32m / s) to the physical entity PLC control system through the OPC UA protocol, with a synchronization delay ≤500ms.

[0096] 6. Twin model self-optimization

[0097] Operation-effect correlation database: records this operation (instruction "speed up 10%") and physical feedback data (actual efficiency increased by 7.5%, energy consumption increased by 3%, operation time 45 seconds);

[0098] Recommended model training: using deep reinforcement learning (DQN algorithm) to train the model, the reward function is:

[0099] R = 0.4 · Δη + 0.3 · (1 - Δη) - 0.3 · t

[0100] Wherein, Δη = 7.5%, Δη = 3%, t = 45 seconds, R = 0.4 x 0.075 + 0.3 x (1 - 0.03) - 0.3 x 45 = -13.3 (negative reward indicates optimization operation time-consuming) is calculated; When the "capacity improvement" scenario appears again, the model preferentially recommends the historical highest reward scheme (such as "adjust the welding current + speed up 5%", R = 2.1).

[0101] The beneficial effects of this embodiment are:

[0102] Through multi-source data fusion, natural interaction interface and self-optimization mechanism, high-precision mapping and efficient interaction of the digital twin model of the automobile production workshop are realized. The actual measurement shows that the production efficiency optimization decision time is shortened by 40%, the equipment utilization is improved by 12%, and the energy consumption is reduced by 5%, verifying the practicality and advancement of the present application.

[0103] The above is only an embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the present application specification, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. An interactive digital twin augmented modeling method, characterized in that: Includes the following steps: Step 1: Construct a real-time dynamically updated digital twin model. Integrate heterogeneous data from multiple sources through a multi-source data interface to form a multi-dimensional twin data layer. The digital twin model synchronizes data with the physical entity through edge computing nodes, and the synchronization frequency is dynamically adjusted according to the entity's state change rate. Step 2: Build a natural interaction interface system that includes a voice interaction module, a gesture recognition module, and a haptic feedback module; Step 3: Receive the operation instructions input by the user through the natural interaction interface, and convert the operation instructions into parameter adjustment instructions that can be executed by the twin model; Step 4: Simulate and calculate the parameter adjustment instructions based on the finite state machine model, and output the adjusted resource load distribution, production cycle change, energy consumption fluctuation trend and potential conflict warning; Step 5: Feed back the simulation calculation results in a three-dimensional visualization manner, and at the same time output the virtual physical properties of the manipulated object through the tactile feedback module. Then, based on the user's confirmation command, synchronize the optimized parameters to the physical entity control system.

2. The interactive digital twin enhanced modeling method according to claim 1, characterized in that: The construction of the multi-dimensional twin data layer in step 1 includes: A spatiotemporal indexing mechanism is used to associate and store multi-source data. The spatiotemporal index uses the physical entity ID as the primary key and associates timestamps, spatial coordinates and attribute data. A data quality assessment model is established to score the integrated data for completeness, accuracy, and timeliness. Low-quality data is repaired using a generative adversarial network (GAN)-based imputation algorithm. The imputation formula is as follows: Where G is the generator network, z is the random noise vector, and x is the random noise vector. bos For incomplete observation data, This is the data after repair.

3. The interactive digital twin enhanced modeling method according to claim 1, characterized in that: The implementation of the voice interaction module in step 2 includes: An industry terminology dictionary is built based on a bidirectional long short-term memory network (Bi-LSTM) to perform noise reduction, word segmentation and intent recognition on the input speech signal, with an intent recognition accuracy of no less than 92%. A contextual understanding model is established, which uses an attention mechanism to associate the current instruction with the historical dialogue sequence, supports instruction completion across rounds, and automatically generates follow-up questions when the instruction is missing key parameters.

4. The interactive digital twin enhanced modeling method according to claim 1, characterized in that: The gesture recognition module described in step 2 captures the three-dimensional coordinates and movement trajectory of the hand using a depth camera, enabling the grasping, moving, and scaling operations of virtual objects in the twin model.

5. The interactive digital twin enhanced modeling method according to claim 1, characterized in that: The operation instruction conversion in step 3 includes: An operation instruction mapping library is constructed to convert natural language instructions and gesture trajectories into standardized control instructions. The standardized control instructions adopt JSON format and include operation object ID, target parameter value, execution priority and time constraint. Ambiguity is resolved for ambiguous commands by comparing the current state of the twin model with historical operation records to determine the user's true intent.

6. The interactive digital twin enhanced modeling method according to claim 1, characterized in that: The simulation calculations described in step 4 include: A discrete event simulation engine is used to simulate the production process refactoring instructions, calculate the waiting time of each process, equipment utilization rate and bottleneck process location, and output the production efficiency optimization range. Based on digital thread technology, the system tracks product lifecycle data and simultaneously calculates the impact chain on subsequent processing steps when equipment parameters are adjusted, generating a heat map of the impact range. The time step of the simulation calculation is dynamically adjusted according to the complexity of the operation. Simple operations use 1 second / step, while complex process reconstruction uses 0.1 seconds / step.

7. The interactive digital twin enhanced modeling method according to claim 1, characterized in that: The 3D visualization feedback mentioned in step 5 includes: The twin model is rendered using Level of Detail (LOD) technology. When the user focuses on a specific area, the model accuracy of that area is automatically improved to the millimeter level, while non-focused areas are simplified to low-polygon models. The key indicators before and after the adjustment are compared and visualized. Dynamic bar charts are used to show changes in production efficiency, color gradients are used to mark differences in resource load, and particle effects are used to simulate changes in material flow paths.

8. The interactive digital twin augmented modeling method according to claim 1, characterized in that: The interactive digital twin enhanced modeling method also includes a twin model self-optimization step: Record user interaction operations and corresponding physical entity feedback data to build an operation-effect association database; A recommendation model is trained using deep reinforcement learning algorithms. When a user repeats similar operation scenarios, the model automatically recommends the best historical adjustment scheme. The reward function of the recommendation model is: R=α·Δη+β·(1-Δ∈)-γ·t Where Δη is the production efficiency improvement rate, Δ∈ is the energy consumption growth rate, t is the operation time, and α, β, and γ are weighting coefficients.

9. The interactive digital twin augmented modeling method according to claim 1, characterized in that: The natural interaction interface system also includes an eye-tracking module. The eye-tracking module captures the user's gaze point through an infrared camera. When the gaze stays on a virtual object for more than a preset time (configurable from 0.5 to 2 seconds), the object is automatically highlighted and its key parameters are displayed.

10. The interactive digital twin enhanced modeling method according to claim 1, characterized in that: The parameter adjustment instructions mentioned in step 3 include resource configuration changes, production process restructuring, and equipment parameter modification.

Images