Digital twin model construction system and method based on UE4

Abstract

The invention discloses a UE4-based digital twin model construction system and method, and relates to the technical field of computer vision, and the system comprises a geometric modeling unit which generates three-dimensional model data based on geometric modeling data of a target physical entity, and achieves dynamic LOD rendering level switching through virtualization processing; the real-time data interaction unit is used for analyzing the real-time sensor data flow and generating equipment state data in combination with a transmission operation and maintenance instruction; the simulation engine unit is used for executing rigid body motion and collision simulation based on the equipment state data and outputting a collision event sequence and state transition data by utilizing an event trigger; the machine learning unit is used for training an LSTM (Long Short Term Memory) model based on the collision event sequence and the state transition data, and optimizing state covariance parameters through adaptive Kalman filtering; and the visual interaction unit is used for realizing data synchronization of the command large screen and the mobile terminal through multi-screen collaboration. According to the method, the digital model capable of accurately reflecting physical entity behaviors and characteristics is constructed.

Description

Technical Field

[0001] This application relates to the field of computer vision technology, and in particular, to a digital twin model construction system and method based on UE4. Background Art

[0002] With the rapid development of intelligent manufacturing, building a digital twin model that can accurately reflect the behavior and characteristics of physical entities has become a core requirement in the fields of intelligent manufacturing, industrial operation and maintenance, etc.

[0003] Traditional digital twin systems mostly rely on static CAD / BIM models, which can only describe the geometric shape of objects and lack the ability to dynamically represent physical properties. Moreover, existing digital twin systems mostly adopt an offline batch processing mode, which is difficult to meet the dynamic data update requirements in real-time interaction scenarios. Therefore, there is a defect that the existing digital twin models cannot accurately reflect the behavior and characteristics of physical entities.

[0004] Therefore, how to build a digital model that can accurately reflect the behavior and characteristics of physical entities has practical application value and significance. Summary of the Invention

[0005] In order to build a digital model that can accurately reflect the behavior and characteristics of physical entities, this application provides a digital twin model construction system and method based on UE4.

[0006] In a first aspect, the invention object of this application is realized by adopting the following technical solutions: A digital twin model construction system based on UE4, comprising: A geometric modeling unit that generates three-dimensional model data including polygon meshes, topological structures, and UV coordinate systems based on the geometric modeling data of the target physical entity, and realizes dynamic LOD rendering level switching through virtualization processing; A real-time data interaction unit that analyzes real-time sensor data streams, combines transmission operation and maintenance instructions, and generates device status data through Kalman filtering; A simulation engine unit that performs rigid body motion and collision simulations based on the device status data, generates environmental special effects, and outputs a collision event sequence and state transition data using event triggers; A machine learning unit that trains an LSTM model based on the collision event sequence and state transition data, and optimizes state covariance parameters through adaptive Kalman filtering; A visualization interaction unit that renders a three-dimensional scene, generates a device health heat map, and realizes data synchronization between the command large screen and mobile terminals through multi-screen collaboration.

[0007] By adopting the above technical solution, UE4 is the abbreviation of Unreal Engine 4, that is, the Unreal Engine 4, which can achieve cross-platform real-time 3D creation; provide a closed-loop architecture based on simulation, learning, visualization and real-time simulation to realize the self-optimization technology of the digital twin system, which is beneficial to constructing a digital model that can accurately reflect the behavior and characteristics of physical entities; specifically, the three-dimensional model data serves as the underlying data source, providing a spatial anchor for real-time data interaction. The switching of the LOD rendering levels generated by virtualization processing directly affects the rendering load of real-time data interaction. The device status data includes sensor data and operation and maintenance instructions, which can drive the physical parameters of the simulation engine to realize real-time data correction of the simulation initial conditions. Through Kalman filtering for noise suppression data preprocessing, the simulation convergence speed is improved; the collision event sequence and state transition data are used as training samples for machine learning, and the synthetic data generated by the simulation environment makes up for the scarcity of real data (such as simulation of rare fault scenarios). Combining the adaptive Kalman filter parameter optimization to reversely improve the simulation accuracy; since the model convergence degree data determines the confidence weight of visual rendering, the LSTM prediction result generates a health prediction heat map (such as visualization of the remaining life of the device), which not only improves the accuracy of device fault prediction, but also is beneficial to real-time debugging feedback through the visual interface. The multi-screen collaboration module reversely transmits the control instructions of the mobile terminal to the real-time data interaction unit, and through the two-way driven data flow, the driving function determines the reliability of the collection modeling data binding (such as sensor position mapping, collision area definition), thereby improving the model accuracy of the digital twin model.

[0008] In a preferred example of the present application: The geometric modeling unit further includes: Obtain basic geographic information data, process the basic geographic information data to obtain a first modeling data set and a second modeling data set; According to the first modeling data set and the second modeling data set, obtain environmental factor data related to the digital twin model, wherein the environmental factor data includes first environmental factor data and second environmental factor data; According to the environmental factor data, obtain a modeling influence factor representing the model accuracy level, wherein the modeling influence factor includes a first modeling influence factor and a second modeling influence factor; According to the modeling influence factor, divide the first modeling data set and the second modeling data set; Obtain the division result, and according to the division result, obtain real-time dynamic parameters.

[0009] By adopting the above technical solutions, it is beneficial to improve the efficiency and accuracy of model construction. Through hierarchical and classified processing of basic geographic information data, this application can effectively distinguish the modeling requirements of different regions, improving the construction efficiency and spatial expression accuracy of the digital twin model. This application introduces modeling influence factors, enabling the model to adjust the level of detail and data density according to the actual application scenario, enhancing the model's adaptability to complex environmental changes. Moreover, by obtaining real-time dynamic parameters from the division results, it provides a basis for dynamic updates for subsequent model optimization, rendering strategy adjustment, and state prediction, improving the intelligent level of the system.

[0010] In a preferred example of this application: The obtaining of the basic geographic information data and the processing of the basic geographic information data to obtain the first modeling data set and the second modeling data set include: Obtaining high-precision terrain data through UAV oblique photography and obtaining BIM data of structures through laser scanning; Performing spatial registration on the terrain data and the BIM data to establish a unified coordinate system; According to the unified coordinate system, obtaining the first modeling data set and the second modeling data set.

[0011] By adopting the above technical solutions, UAV oblique photography is used to obtain terrain data, laser scanning is used to obtain BIM structure data, and spatial registration is performed to establish a unified coordinate system, thereby generating two modeling data sets. This application combines UAV oblique photography and laser scanning technologies to achieve high-precision data acquisition of macroscopic terrain and microscopic structures, providing a reliable data source for high-quality three-dimensional modeling. By establishing a unified coordinate system through spatial registration, it ensures the consistency of multi-source heterogeneous data in spatial positions, avoiding model misalignment and distortion problems.

[0012] In a preferred example of this application: According to the environmental element data, obtaining modeling influence factors representing the model accuracy level, including: Obtaining preset modeling specification parameters, and screening the first environmental element data according to the preset modeling specification parameters to obtain a first modeling parameter set, where the first modeling parameter set includes a first geological structure parameter, a first building structure parameter, and a first material texture parameter; According to the first modeling parameter set, obtaining a first modeling influence factor; According to the preset modeling specification parameters, screening the second environmental element data to obtain a second modeling parameter set, where the second modeling parameter set includes a second hydrometeorological parameter, a second traffic network parameter, and a second sensor parameter; According to the second modeling parameter set, obtaining a second modeling influence factor.

[0013] By adopting the above technical solutions, it is beneficial to precisely control the accuracy level of the model. In this application, key modeling parameters are screened out through modeling specification parameters, enabling scientific evaluation of the modeling priorities and accuracy requirements in different regions and realizing refined hierarchical modeling. The first and second modeling influence factors respectively correspond to the fine structure and the lightweight region, enabling the model to not only meet the high-fidelity display of the core region but also take into account the overall performance and visualization fluency.

[0014] In a preferred example of this application: The division of the first modeling data set and the second modeling data set according to the modeling influence factor includes: According to the first modeling influence factor, a fine model partition at the LOD500 level is established, and based on the fine model partition, the first modeling data set is divided; According to the second modeling influence factor, a lightweight model partition at the LOD200 level is established, and based on the lightweight model partition, the second modeling data set is divided.

[0015] By adopting the above technical solutions, the fine and lightweight model partitions are divided, effectively balancing the model accuracy and the consumption of computing resources, and being suitable for efficient rendering under different terminal devices and interaction scenarios. In this application, on the premise of ensuring the visual quality, the model complexity is reasonably configured, reducing the graphics rendering pressure and improving the operation efficiency and stability of the system.

[0016] In a preferred example of this application: The device status data is a device status matrix including time series characteristics, including: Among them, is the fusion status matrix; is the adaptive weight coefficient (0.6 ≤ ≤ 0.9); is the original sensor data; is the transmission and operation and maintenance instruction set.

[0017] By adopting the above technical solutions, the integrity and accuracy of the device status description are enhanced. In this application, by constructing a status matrix including sensor data and operation and maintenance instructions, the operating status of the device and its external control behavior can be comprehensively reflected. The use of the adaptive weight coefficient (0.6 ≤ ≤ 0.9) can dynamically adjust the data fusion ratio under different working conditions, making the state estimation closer to the actual operating conditions and improving the system perception accuracy.

[0018] In a preferred example of this application: The simulation engine unit executes a collision detection algorithm based on rigid body dynamics. When a collision event occurs: Call the UE4 Niagara system to generate hydrodynamic special effects; Output state transition data containing the collision force vector through an event generator; Input the collision time series into the LSTM prediction model to dynamically optimize the state covariance matrix: Wherein, is the estimated covariance matrix, which is a symmetric matrix; is the prior covariance matrix; is the state transition matrix; is the transpose matrix of; is the process noise covariance matrix.

[0019] By adopting the above technical solutions, the UE4 Niagara system generates hydrodynamic special effects, enhancing the realism and interactive experience of the digital twin scenario; this application outputs state transition data through an event trigger, achieving real-time response to sudden physical events, improving the system's recognition ability for abnormal working conditions. This application combines the LSTM model with the covariance matrix update formula to dynamically adjust the Kalman filter parameters, effectively suppressing noise interference and enhancing the stability and prediction accuracy of state estimation.

[0020] In the second aspect, the invention object of this application is realized by adopting the following technical solutions: A method for constructing a digital twin model based on UE4, the method includes: Utilize a geometric modeling unit to generate three-dimensional model data containing polygon meshes, topological structures, and UV coordinate systems based on the geometric modeling data of the target physical entity, and perform virtualization processing on the three-dimensional model data to achieve dynamic LOD rendering level switching; Utilize a real-time data interaction unit to parse real-time sensor data streams, and combine transmission operation and maintenance instructions to generate device state data through Kalman filter processing; Utilize a simulation engine unit to perform rigid body motion and collision simulations based on the device state data, and generate environmental special effects, and output a collision event sequence and state transition data through an event trigger; Utilize a machine learning unit to train an LSTM model based on the collision event sequence and state transition data, and optimize the state covariance parameters through adaptive Kalman filtering; Utilize a visual interaction unit to render a three-dimensional scene, and generate a device health heat map, and realize data synchronization between the command large screen and the mobile terminal through multi-screen collaboration technology.

[0021] In a preferred example of this application: the modeling steps of the geometric modeling unit include: Obtain basic geographic information data; Process the basic geographic information data to obtain a first modeling data set and a second modeling data set; Obtain environmental factor data related to the digital twin model according to the first modeling data set and the second modeling data set, where the environmental factor data includes first environmental factor data and second environmental factor data; Obtain modeling impact factors representing the model accuracy level according to the environmental factor data, where the modeling impact factors include a first modeling impact factor and a second modeling impact factor; Divide the first modeling data set and the second modeling data set according to the modeling impact factors; Obtain the division result, and obtain real-time dynamic parameters according to the division result for model optimization and update.

[0022] By adopting the above technical solution, the modeling task is decomposed and parameter screening is carried out based on a standardized process to ensure that the modeling process complies with industry specifications and improve the model quality and consistency; through the evaluation of the modeling impact factors, the modeling area and detail level are reasonably divided to avoid resource waste and improve the rendering efficiency and system performance; the present application obtains real-time dynamic parameters according to the division result, enabling the model to have the ability of self-evolution and adapt to the changing physical entity state and business requirements.

[0023] In a third aspect, the invention object of the present application is realized by adopting the following technical solution: A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above method for constructing a digital twin model based on UE4 are realized.

[0024] In summary, the present application includes at least one of the following beneficial technical effects: 1. It covers the full life cycle management from physical entity modeling, data collection, state simulation, intelligent analysis to visual display, forming a complete digital twin closed-loop system. The present application constructs a complete closed-loop digital twin system framework, and through the collaborative work of each module, it is beneficial to improve the intelligence level and system integration degree of the digital twin model construction system; 2. By decomposing the modeling task and screening parameters through a standardized process, it is beneficial to improve the scientificity and standardization of 3D modeling; the present application realizes hierarchical modeling and optimized resource allocation through the evaluation of modeling impact factors. Description of the Drawings

[0025] Figure 1 is a framework diagram of a digital twin model construction system based on UE4 in an embodiment of the present application; Figure 2 is a method step diagram of a geometric modeling unit in a digital twin model construction system based on UE4 in an embodiment of the present application; Figure 3It is a flowchart of a method for constructing a digital twin model based on UE4 in an embodiment of the present application. Detailed implementation manners

[0026] The following further elaborates on the present application in conjunction with the accompanying drawings.

[0027] In one embodiment, as Figure 1 shown, the present application discloses a digital twin model construction system based on UE4. The digital twin model construction system includes a geometric modeling unit, a real-time data interaction unit, a simulation engine unit, a machine learning unit, and a visualization interaction unit. The geometric modeling unit generates three-dimensional model data including polygon meshes, topological structures, and UV coordinate systems based on the geometric modeling data of the target physical entity, and realizes dynamic LOD rendering level switching through virtualization processing. The polygon mesh is a 3D model surface representation composed of vertices, edges, and faces, which determines the geometric shape of the object. The topological structure is used to describe the connection relationship of the object's faces / edges / vertices and is used for collision detection and physical simulation. The UV coordinate system is a 2D texture mapping coordinate that determines the texture mapping position of the material on the 3D model surface. Dynamic LOD is to switch the model detail level in real time according to the viewing distance / performance requirements.

[0028] The real-time data interaction unit analyzes the real-time sensor data stream, combines the transmission and operation and maintenance instructions, and generates device status data through Kalman filtering. The real-time data interaction unit uses the Modbus TCP protocol to interact with the sensors for detecting the device status, analyzes the Modbus TCP protocol, extracts the device status code (such as the fault code 0x8001), and obtains the transmission and operation and maintenance instructions through the OPC-UA protocol (in actual applications, for the abnormal status of the device, an abnormal threshold for triggering the transmission and operation and maintenance instructions needs to be set. Different types or categories of device abnormal statuses need to set corresponding thresholds respectively to immediately detect the abnormal status of the device and trigger the transmission and operation and maintenance instructions at the first time, such as the set value of the shutdown threshold).

[0029] Specifically, the implementation steps of the Kalman filtering of the real-time data interaction unit include: Based on the state equation: , where and are the n-dimensional vectors at the k-th moment and the (k - 1)-th moment respectively, including all key state variables describing the dynamic characteristics of the system, such as mechanical systems (position, speed, acceleration, and angle, etc.), device monitoring systems (temperature, vibration amplitude, rotation speed, and fault flag bits, etc.), and digital twin systems (physical entity coordinates, material status, and environmental parameters, etc.); is the state transition matrix (including the time step Δt = 0.1s); is the control input matrix, which converts the control instruction into a gain matrix for state change; is the operation and maintenance instruction vector; is the process noise; configure the noise covariance matrix , identifying the process noise variances for two state variables (such as position, state); high value indicates low model confidence, and the filter relies more on sensor measurements; low value indicates high model confidence, and the filter trusts the predicted values more; configure the measurement noise covariance , high value, indicating low sensor accuracy, and the filter relies more on model predictions, and vice versa indicating high sensor accuracy; use the Eigen library for matrix operations, with an update period of 100 ms.

[0030] Furthermore, the device status data is a device status matrix containing temporal features, including: Among them, is the fusion status matrix; is the adaptive weight coefficient (0.6 ≤ ≤ 0.9); is the raw sensor data; is the transmission operation and maintenance instruction set.

[0031] The simulation engine unit performs rigid body motion and collision simulations based on the device status data, generates environmental special effects, and outputs a collision event sequence and state transition data using an event trigger; rigid body dynamics is used to study the motion laws of objects under the action of forces (ignoring deformation); collision detection is used to determine whether geometric penetration occurs between objects; an event trigger refers to generating an asynchronous notification when a specific condition is detected; specifically, the physical simulation uses the UE4 Chaos physics system and enables SIMD acceleration; the integration algorithm for rigid body motion simulation uses the Verlet integration method, with a time step Δt = 0.016 s (60 FPS); the collision detection frequency is to perform two consecutive collision detections per frame, convert the CAD model to Convex Collision (convex hull simplification rate ≥ 85%), and use the Sphere Sweep algorithm to generate approximate collision bodies for complex curved surfaces. For example, a car bumper collision body consists of 32 convex hulls, and the total number of patches ≤ 512; the constraint solving algorithm uses the PGS (Progressive Gradient Descent) algorithm to solve contact constraints, with an error tolerance of 1e-5.

[0032] Furthermore, the simulation engine unit executes a collision detection algorithm based on rigid body dynamics. When a collision event occurs: Invoke the UE4 Niagara system to generate fluid dynamics special effects; output state transition data containing collision force vectors through an event generator; input the collision time series into the LSTM prediction model to dynamically optimize the state covariance matrix: Among them, is the estimated covariance matrix, which is a symmetric matrix; is the prior covariance matrix; is the state transition matrix; is the transpose matrix of; is the process noise covariance matrix.

[0033] When a collision event occurs, the event trigger determines the collision time when the contact force > 1000N. At this time, the coordinates of the collision point are recorded (accuracy ±1cm) and an event in JSON format is output.

[0034] Among them, the generation of environmental special effects uses a particle system. The particle system generates an explosion effect based on parameters of 2000 particles per second and a life cycle of 0.5 - 2s. The environmental special effects also include smoke simulation, which uses the SPH (Smoothed Particle Hydrodynamics) algorithm for smoke simulation, and the collision event is associated with the Wwise audio engine for sound effect simulation.

[0035] The machine learning unit trains the LSTM model based on the collision event sequence and state transition data, and optimizes the state covariance parameters through adaptive Kalman filtering; adaptive Kalman filtering is a filtering algorithm that dynamically adjusts the noise parameters according to the system state, uses the recursive least squares method (RLS) to update the noise statistics, and the update period is 5s; the time window length of the collision event sequence is 10s, and the sampling rate is 100Hz; the state transition data includes device mode switch records; the loss function of the LSTM model is the cross - entropy loss function, the optimizer is Adam (lr = 0.001), Epochs is 100, and early stopping is used to monitor the validation set loss.

[0036] The visualization interaction unit renders the 3D scene and generates a device health heat map, and realizes data synchronization between the command large screen and the mobile terminal through multi - screen collaboration; the device health heat map maps the device state parameters into a color gradient map, and multi - screen collaboration is the synchronous display of data between the command large screen and the mobile terminal; specifically, the heat map maps the device health index HPI (0 - 100) into the HSV color space, (H = 0°~360°), uses bilinear interpolation to generate smooth gradients, and the WebGL shader updates the heat map every frame (frame rate ≥ 30FPS); among them, multi - screen collaboration is realized based on the data synchronization protocol, that is, uses WebSocket to transmit JSON data (compression rate ≥ 60%), and combines the client prediction algorithm for latency compensation.

[0037] Further, the calculation formula for the device health index HPI is: , where is the number of parameters participating in the calculation, usually corresponding to the number of key monitoring indicators of the device; is the parameter index; is the parameter reference value (ideal state or maximum allowable value), which is a reference standard for measuring the current state; is the current measured value of the i-th parameter, reflecting the real-time operating state of the device.

[0038] In one embodiment, as Figure 2 shown, the geometric modeling unit further includes the following method steps: S1: Obtain basic geographic information data, process the basic geographic information data, and obtain a first modeling data set and a second modeling data set.

[0039] In this embodiment, the basic geographic information data includes digital data of terrain and landform features, such as drone images, satellite images, DEM (Digital Elevation Model), and DOM (Digital Orthophoto Map).

[0040] Specifically, step S1 includes: S11: Obtain high-precision terrain data through drone oblique photography and obtain BIM data of structures through laser scanning.

[0041] Specifically, the flight route setting of the drone includes an orthophoto route (height 120m) and an oblique route (height 100m, side overlap rate 70%); the scanning strategy of laser scanning includes structure scanning (point cloud density ≥ 1000 points / m²) and terrain scanning (point cloud density ≥ 200 points / m²).

[0042] S12: Perform spatial registration on the terrain data and BIM data to establish a unified coordinate system.

[0043] Specifically, first use the ORB (Oriented FAST and Rotated BRIEF) algorithm to extract key points and extract image features, and then perform point cloud feature matching. The unified coordinate system includes the layout of control points (more than 10) arranged on ground control points (GCP), and then the seven-parameter Bursa-Wolf model is used for coordinate transformation. The calculation formula for the transformation parameters is: where , , are the target coordinate parameters after transformation; , , are the original coordinate parameters before transformation; ΔX, ΔY, and ΔZ are translation parameters, k is the scale parameter, and c is the rotation parameter.

[0044] S13: Obtain the first modeling data set and the second modeling data set according to the unified coordinate system.

[0045] Specifically, the first modeling data set corresponds to the grading standard of the LOD500-level fine model data set, which is applicable to core buildings (such as the reactor building of a nuclear power plant), with a point cloud density ≥ 5000 points / m² and a texture resolution ≥ 5mm / pixel; the second modeling data set corresponds to the grading standard of the LOD200-level lightweight model data set, which is suitable for auxiliary facilities (such as factory roads and green belts) areas, and the data requires a point cloud density ≥ 500 points / m² and a texture resolution ≥ 20mm / pixel.

[0046] In this embodiment, the optimization of the data set and the processing of abnormal data include: using a Voxel Grid filter (voxel size 0.1m) for noise reduction to perform statistical filtering to remove abnormal points (standard deviation threshold 3σ), then generating a triangular mesh using the Poisson surface reconstruction algorithm, optimizing the mesh using Laplacian smoothing, and in the texture mapping step, using Spherical Mapping to project the image onto the mesh surface and performing texture stitching based on seamless fusion of SIFT feature matching to improve the modeling accuracy.

[0047] Furthermore, the Icp-SVD improved algorithm can be used for multi-source data spatial registration to improve the registration accuracy: first add a normal vector constraint term, and the registration success rate is improved. The calculation formula is: , where R is the rotation matrix, represents a rigid body transformation, that is, only rotating and translating operations are used to transform the point cloud, keeping its geometric structure unchanged; j is the point cloud data index; is the jth three-dimensional coordinate point in the target point cloud; is the jth three-dimensional coordinate point in the source point cloud; R is a 3X3 rotation matrix used to rotate the source point cloud to the posture aligned with the target point cloud, and the rotation matrix needs to satisfy the orthogonality (R^T R = I) and the condition that the determinant is 1; is a 3×1 translation vector, such as ; N is the number of point pairs participating in the registration.

[0048] S2: Obtain the environmental element data related to the digital twin model according to the first modeling data set and the second modeling data set, where the environmental element data includes the first environmental element data and the second environmental element data.

[0049] In this embodiment, the first environmental element data reflects the environmental parameters of the regional static natural attributes, including geological structure, terrain features, soil types, etc.; the second environmental element data includes the environmental parameters of dynamic changes and human interventions, such as hydrological conditions, vegetation coverage, meteorological disaster risks, etc.

[0050] Specifically, the geological structure is sourced from the regional geological investigation report; when extracting the terrain features, first fuse the data features of DEM data (30m resolution) and UAV oblique photography DOM (0.5m resolution), and use GIS hydrological analysis tools to extract areas with a slope of ≥25° for slope grading, and calculate the tributary density based on morphological operations to obtain the terrain valley density. The hydrological conditions are sourced from the sensor monitoring data of water level gauges, flow meters, and velocity profilers arranged at key cross-sections of the river; fuse and register the multi-source environmental element data: that is, calibrate the sensor clocks using the NTP protocol for time synchronization, and perform spatial registration through GCP control points to unify the spatio-temporal reference of the multi-source data. And perform data interpolation and parameter standardization processing on the meteorological data and ecological data. The standardization of meteorological data includes converting the meteorological data into the CF (Climate and Forecast) standard format, and the standardization of ecological data includes mapping the ecological data to the ecological red line evaluation index system (GB 35571-2017).

[0051] S3: According to the environmental element data, obtain the modeling influence factors representing the model accuracy level, where the modeling influence factors include the first modeling influence factor and the second modeling influence factor.

[0052] In this embodiment, step S3 includes: S31: Obtain the preset modeling specification parameters, and according to the preset modeling specification parameters, screen the first environmental element data to obtain the first modeling parameter set, where the first modeling parameter set includes the first geological structure parameter, the first building structure parameter, and the first material texture parameter.

[0053] In this embodiment, the preset modeling specification parameters are data screening rules formulated according to the "Technical Standard for the Urban Information Model (CIM) Foundation Platform" (GB / T 38667-2020), including the accuracy level, coordinate system (CGCS2000), and semantic label specification.

[0054] Specifically, obtain the specification document, convert the accuracy level in the specification document into the StaticMeshLOD grouping strategy of UE4 (e.g., LOD0 corresponds to a 1-cm accuracy model), and screen and extract the first environmental elements. The first geological structure parameters are a set of physical property parameters describing the geological structure of the target area, including rock and soil types (soil / rock layer distribution obtained through geological borehole data (such as clay, sandstone, gravel layer)), geological age (refers to the tectonic activity grading based on the seismic intensity zoning map (such as Quaternary sedimentary layer, Cretaceous bedrock)), foundation bearing capacity (foundation bearing index calibrated through static cone penetration test (CPT) or standard penetration test (SPT)), and groundwater level (extracted by combining the aquifer depth and permeability coefficient of the hydrogeological model). The first building structure parameters are a structured data set characterizing the physical properties of artificial structures, including building types (building function classification parsed through BIM data (residential / commercial / industrial)), structural systems (such as load-bearing system parameters of frame structure / shear wall structure / steel structure, etc.), material properties (concrete grade such as C30 / C40, steel yield strength such as Q235 / Q355), and geometric topology (coordinates of beam-column joints, floor thickness, door and window opening sizes).

[0055] S32: Obtain the first modeling influence factor according to the first modeling parameter set.

[0056] In this embodiment, the first modeling influence factor is the core parameter controlling the geometric model accuracy, including data integrity (missing rate ≤ 5%), parameter accuracy (coordinate error ≤ 0.5 m), and topological consistency (the area cannot overlap ≤ 0.1%).

[0057] Specifically, use the PDAL library for coordinate transformation (WGS84 → CGCS2000), and adopt the spatial registration method of the ICP algorithm to achieve point cloud registration; among them, the data integrity is scored and calculated using the point cloud data integrity of geological / building / material data to obtain the data missing rate , when the data missing rate is greater than 5%, the data integrity verification is to use CloudCompare for point cloud density analysis: , where is the total amount of unprocessed point cloud data collected originally (unit: number of points); is the amount of effective point cloud data actually used in the UE4 model (unit: number of points). If the data missing rate > 5%, trigger the data re-collection process (by re-flying the drone or re-scanning with lidar).

[0058] The parameter accuracy is verified using the coordinate fusion algorithm based on the Kalman filter: where is the state estimate value at time k (such as coordinate position, velocity, acceleration); is the Kalman gain; is the original sensor observation value at time k; is the state prediction value based on time k−1; the standard deviation of the output coordinate error σ ≤ 0.3m (confidence level 95%).

[0059] Topological consistency detection can be applied to the CGAL library for surface overlap detection, and the output results are normalized.

[0060] The calculation formula of the first modeling influence factor is: , where is the first modeling influence factor; D is data integrity (missing rate ≤ 5%); P is parameter accuracy (coordinate error ≤ 0.5m); T is topological consistency (surface overlap ≤ 0.1%). 、 、 are the weights determined by the AHP (Analytic Hierarchy Process) method (example: = 0.42, = 0.35, = 0.23).

[0061] S33: According to the preset modeling specification parameters, filter the second environmental element data to obtain the second modeling parameter set, where the second modeling parameter set includes second hydrometeorological parameters, second traffic network parameters, and second sensor parameters.

[0062] In this embodiment, the second hydrometeorological data is output by the SWAT model (time resolution 15min), the second traffic network data extracts the road center line through OpenStreetMap (accuracy ≤ 1m), and the second sensor data transmits the vibration monitoring value through the LoRaWAN protocol (frequency 10Hz).

[0063] Specifically, the second hydrometeorological parameters are a set of dynamic environmental data describing the hydrological cycle and meteorological characteristics of the target area, including rainfall, river water level (river water level change curve based on the hydrological model), wind speed and direction (near-surface wind field data with a 10-minute resolution), and snow cover (winter snow thickness and distribution range). The second traffic network parameters are a set of dynamic parameters characterizing the real-time operation state of the road network, including vehicles, average speed, signal light timing, and accident hotspots. The second sensor parameters are a set of engineering parameters used to describe the physical characteristics and deployment scheme of the monitoring equipment, including installation coordinates, sampling frequency, detected data type, range, and data transmission standard.

[0064] S34: Obtain the second modeling influence factor according to the second modeling parameter set.

[0065] Specifically, the second modeling impact factor is a parameter set that affects the accuracy of dynamic simulation, including real-time performance (data update frequency ≥ 1 Hz), coverage (sensor blind area ≤ 10%), and noise ratio (signal-to-noise ratio ≥ 20 dB).

[0066] Specifically, , where, is the second modeling impact factor; R is real-time performance (data update frequency ≥ 1 Hz), C is coverage (sensor blind area ≤ 10%), and N is noise ratio (signal-to-noise ratio ≥ 20 dB). 、 and are weights determined by the grey relational analysis method (example: = 0.38, = 0.32, = 0.30).

[0067] In this embodiment, the real-time performance evaluation is carried out by using the end-to-end delay test: T_delay = T_sensor + T_edge + T_cloud, where T_delay is the end-to-end delay time; T_sensor is the sensor data acquisition time; T_edge is the edge computing node data processing time; T_cloud is the cloud data transmission and processing time.

[0068] The coverage is analyzed and quantified by using the Voronoi diagram coverage rate: , where, is the area effectively covered by the sensor; is the geometric area of the target area; the blind area is identified by using the DBSCAN clustering algorithm to identify the weak signal area (spatial distance threshold = 5 m, minimum number of points in the core point neighborhood = 5) to determine the detection blind area. The QoS hierarchical transmission is implemented through the MQTT protocol (traffic data priority P = 2, meteorological data P = 1) S4: Divide the first modeling data set and the second modeling data set according to the modeling impact factor.

[0069] In this embodiment, step S4 includes: S41: Establish a LOD500-level fine model partition according to the first modeling impact factor, and divide the first modeling data set according to the fine model partition.

[0070] In this embodiment, the LOD500-level fine model partition is a model area that meets the 1:100 scale accuracy requirement (such as precision mechanical parts, geological fault zones), the number of polygon patches ≥ 10^6, and includes detailed information such as material textures and normal maps.

[0071] Specifically, according to the first modeling impact factor score, the LOD500-level fine model area is divided. A dynamic partitioning threshold is set to divide fine models of different levels of refinement, where the dynamic partitioning threshold is a partitioning boundary parameter automatically adjusted based on the modeling impact factor score. For example, areas with a geological structure integrity score ≥ 0.8 are included in the LOD500 partition.

[0072] S42: According to the second modeling impact factor, establish the LOD200-level lightweight model partition, and divide the second modeling data set according to the lightweight model partition.

[0073] Specifically, according to the second modeling impact factor score, divide the LOD200-level fine model area; the LOD200-level lightweight model partition satisfies the model area with a scale accuracy requirement of 1:500 (such as urban roads and building outlines), the number of polygon patches ≤ 10^4, and the mesh simplification algorithm (such as Quadric Error Metrics) is used; in actual applications, a dynamic adjustment mechanism and high-priority areas of the digital twin model also need to be set up to achieve model lightweight while improving the model refinement. High-priority areas such as areas with a geological structure score ≥ 0.8 of the mountain are forced into the LOD500 partition; the dynamic adjustment mechanism is such that when the sensor detects geological activities (such as the seismic waveform amplitude > the threshold), the LOD level of the local area is automatically increased.

[0074] In this embodiment, a partition conflict arbitration also needs to be set up. When the boundaries of the LOD500 and LOD200 partitions are inconsistent, start the following arbitration rules: when the first modeling impact factor score is greater than 0.7, it is included in the LOD500-level fine model; when the second modeling impact factor score is less than 0.4, it is included in the LOD200-level fine model; otherwise, set up an LOD300-level fine model to model the partition conflict area, and trigger a partition modeling instruction according to the modeling information and conflict situation of the partition conflict area to prompt the user for modeling details.

[0075] S5: Obtain the partitioning result, and according to the partitioning result, obtain the real-time dynamic parameters.

[0076] In this embodiment, the partitioning result is the model partition result (LOD500 / LOD200); the real-time dynamic parameters are a set of system parameters dynamically adjusted according to the model partition result (LOD500 / LOD200), such as the rendering level switching threshold, sensor fusion weight, and collision detection sensitivity.

[0077] Specifically, dynamic parameter vectors are set, such as sensor weights and filtering coefficients; partition weights are associated with LOD500 and LOD200 respectively. Then, the sensor weight adjustment adopts sensor weighted fusion based on partition priority. For example, sensors in high-precision areas are given higher weights. The adjustment content of the Kalman filter covariance is as follows: the process noise covariance for high-precision requirements is reduced, and the process noise covariance for low-precision requirements is increased. The rendering level switching threshold is calculated using the dynamic LOD switching distance threshold according to priority: the higher the priority (LOD500), the closer the switching distance (precision first).

[0078] In one embodiment, as Figure 3 shown, a method for constructing a digital twin model based on UE4 is provided. This method for constructing a digital twin model based on UE4 is applied to a system for constructing a digital twin model based on UE4. The method for constructing a digital twin model based on UE4 specifically includes the following steps: S10: Using the geometric modeling unit, generate three-dimensional model data including polygon meshes, topological structures, and UV coordinate systems based on the geometric modeling data of the target physical entity, and perform virtualization processing on the three-dimensional model data to achieve dynamic LOD rendering level switching; S20: Use the real-time data interaction unit to parse the real-time sensor data stream, and combine with the transmission operation and maintenance instructions to generate device status data through Kalman filter processing; S30: Use the simulation engine unit to perform rigid body motion and collision simulations based on the device status data, and generate environmental special effects, and output collision event sequences and state transition data through event triggers; S40: Use the machine learning unit to train the LSTM model based on the collision event sequence and state transition data, and optimize the state covariance parameters through adaptive Kalman filtering; S50: Use the visualization interaction unit to render the three-dimensional scene, and generate a device health heat map, and realize data synchronization between the command large screen and the mobile terminal through multi-screen collaboration technology.

[0079] In this embodiment, the modeling steps of the geometric modeling unit include: Obtain basic geographic information data; Process the basic geographic information data to obtain the first modeling data set and the second modeling data set; According to the first modeling data set and the second modeling data set, obtain environmental element data related to the digital twin model, where the environmental element data includes the first environmental element data and the second environmental element data; According to the environmental element data, obtain modeling influence factors representing the model accuracy level, where the modeling influence factors include the first modeling influence factor and the second modeling influence factor; Divide the first modeling data set and the second modeling data set according to the modeling influencing factors; Obtain the division result, and obtain real-time dynamic parameters according to the division result for model optimization and update.

[0080] It should be understood that the sequence numbers of the steps in the above embodiments do not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0081] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: S10: Use the geometric modeling unit to generate three-dimensional model data including polygon meshes, topological structures, and UV coordinate systems based on the geometric modeling data of the target physical entity, and perform virtualization processing on the three-dimensional model data to achieve dynamic LOD rendering level switching; S20: Use the real-time data interaction unit to parse the real-time sensor data stream, and combine the transmission and operation and maintenance instructions to generate device status data through Kalman filter processing; S30: Use the simulation engine unit to perform rigid body motion and collision simulations based on the device status data, and generate environmental special effects, and output a collision event sequence and state transition data through an event trigger; S40: Use the machine learning unit to train the LSTM model based on the collision event sequence and state transition data, and optimize the state covariance parameters through adaptive Kalman filtering; S50: Use the visual interaction unit to render the three-dimensional scene, and generate a device health heat map, and realize data synchronization between the command large screen and the mobile terminal through multi-screen collaboration technology.

[0082] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0083] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0084] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A digital twin model construction system based on UE4, characterized in that Including: A geometric modeling unit that generates three-dimensional model data including polygon meshes, topological structures, and UV coordinate systems based on the geometric modeling data of the target physical entity, and realizes dynamic LOD rendering level switching through virtualization processing; A real-time data interaction unit that analyzes real-time sensor data streams, combines transmission and operation and maintenance instructions, and generates device status data through Kalman filtering; A simulation engine unit that performs rigid body motion and collision simulations based on the device status data, generates environmental special effects, and outputs a collision event sequence and state transition data using event triggers; A machine learning unit that trains an LSTM model based on the collision event sequence and state transition data, and optimizes state covariance parameters through adaptive Kalman filtering; A visualization interaction unit that renders a three-dimensional scene, generates a device health heat map, and realizes data synchronization between the command large screen and the mobile terminal through multi-screen collaboration.

2. The digital twin model construction system based on UE4 according to claim 1, characterized in that, The geometric modeling unit further includes: Obtaining basic geographic information data, processing the basic geographic information data to obtain a first modeling data set and a second modeling data set; According to the first modeling data set and the second modeling data set, obtaining environmental element data related to the digital twin model, where the environmental element data includes first environmental element data and second environmental element data; According to the environmental element data, obtaining a modeling influence factor representing the model accuracy level, where the modeling influence factor includes a first modeling influence factor and a second modeling influence factor; According to the modeling influence factor, dividing the first modeling data set and the second modeling data set; Obtaining a division result, and according to the division result, obtaining real-time dynamic parameters.

3. A digital twin model construction system based on UE4 according to claim 2, characterized in that, The obtaining basic geographic information data, processing the basic geographic information data to obtain a first modeling data set and a second modeling data set includes: Obtaining high-precision terrain data through drone oblique photography and obtaining structural BIM data through laser scanning; Performing spatial registration on the terrain data and BIM data to establish a unified coordinate system; According to the unified coordinate system, obtaining a first modeling data set and a second modeling data set.

4. A digital twin model construction system based on UE4 according to claim 2, characterized in that According to the environmental element data, obtaining a modeling influence factor representing the model accuracy level includes: Obtaining preset modeling specification parameters, screening the first environmental element data according to the preset modeling specification parameters to obtain a first modeling parameter set, where the first modeling parameter set includes first geological structure parameters, first building structure parameters, and first material texture parameters; According to the first modeling parameter set, obtaining a first modeling influence factor; According to the preset modeling specification parameters, screening the second environmental element data to obtain a second modeling parameter set, where the second modeling parameter set includes second hydrological and meteorological parameters, second traffic network parameters, and second sensor parameters; According to the second modeling parameter set, obtaining a second modeling influence factor.

5. A digital twin model construction system based on UE4 according to claim 2, characterized in that The dividing the first modeling data set and the second modeling data set according to the modeling influence factor includes: According to the first modeling influence factor, establishing a LOD500-level fine model partition, and dividing the first modeling data set according to the fine model partition; Establish an LOD200-level lightweight model partition according to the second modeling influence factor, and divide the second modeling data set according to the lightweight model partition.

6. The digital twin model construction system based on UE4 according to claim 2, wherein The device status data is a device status matrix containing time series characteristics, including: Among them, is the fusion state matrix; is the adaptive weight coefficient (0.6 ≤ ≤ 0.9); is the original sensor data; is the transmission operation and maintenance instruction set.

7. A digital twin model construction system based on UE4 according to claim 1 or 2, characterized in that, The simulation engine unit executes a collision detection algorithm based on rigid body dynamics. When a collision event occurs: Call the UE4 Niagara system to generate fluid dynamics special effects; Output status transition data containing collision force vectors through an event generator; Input the collision time series into the LSTM prediction model to dynamically optimize the state covariance matrix: Among them, is the estimated covariance matrix, which is a symmetric matrix; is the prior covariance matrix; is the state transition matrix; is the transpose matrix of; is the process noise covariance matrix.

8. A method for constructing a digital twin model based on UE4, characterized in that, The method includes: Using the geometric modeling unit, generate three-dimensional model data containing polygon meshes, topological structures, and UV coordinate systems based on the geometric modeling data of the target physical entity, and perform virtualization processing on the three-dimensional model data to achieve dynamic LOD rendering level switching; Use the real-time data interaction unit to parse the real-time sensor data stream, and combine the transmission and operation and maintenance instructions to generate device status data through Kalman filter processing; Use the simulation engine unit to perform rigid body motion and collision simulations based on the device status data, generate environmental special effects, and output a collision event sequence and status transition data through an event trigger; Use the machine learning unit to train the LSTM model based on the collision event sequence and status transition data, and optimize the state covariance parameters through adaptive Kalman filtering; Use the visual interaction unit to render the three-dimensional scene, generate a device health heat map, and realize data synchronization between the command large screen and the mobile terminal through multi-screen collaboration technology.

9. A method for constructing a digital twin model based on UE4 according to claim 8, characterized in that, The modeling steps of the geometric modeling unit include: Obtain basic geographic information data; Process the basic geographic information data to obtain a first modeling data set and a second modeling data set; According to the first modeling data set and the second modeling data set, obtain environmental element data related to the digital twin model, where the environmental element data includes first environmental element data and second environmental element data; According to the environmental element data, obtain a modeling influence factor representing the model accuracy level, where the modeling influence factor includes a first modeling influence factor and a second modeling influence factor; According to the modeling influence factor, divide the first modeling data set and the second modeling data set; Obtain the division result, and obtain real-time dynamic parameters according to the division result for model optimization and update.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method for constructing a digital twin model based on UE4 according to any one of claims 8 to 9.

Images