Lightweight conversion and high-performance real-time loading processing system of large-scale BIM data

Abstract

The application relates to the technical field of building auxiliary design, and discloses a large-scale BIM data lightweight conversion and high-performance real-time loading processing system, which comprises the following units: a data analysis unit for extracting the spatial coordinates, semantic attributes and hierarchical paths of components; an associated dependency analysis unit for identifying reference logic to determine a topological constraint transmission factor; a comprehensive fidelity weight calculation unit for determining a comprehensive weight based on attribute weight and neighborhood weight superposition; a topological feature locking unit for extracting feature points and associating fidelity determination parameters; and an adaptive grid processing unit for executing topological compression and locking feature point coordinates to generate a multi-level precision model; and a spatial retrieval organization unit for constructing a retrieval framework and mapping the model into a spatial tile. The application cooperatively restricts the feature point coordinate locking and the fidelity parameters, so that the interface is consistent when the model level is replaced, and geometric drift is eliminated.

Description

Technical Field

[0001] This invention relates to a lightweight conversion and high-performance real-time loading and processing system for large-scale BIM data, belonging to the field of building auxiliary design technology. Background Technology

[0002] As a data carrier integrating geometry, materials, and business attributes, Building Information Modeling (BIM) plays a core role in the collaboration of the entire project lifecycle. To adapt to the browsing needs of lightweight clients such as web and mobile devices, the industry generally adopts mesh simplification and multi-level detail expression techniques. By reducing the description precision of non-critical components, the consumption of computing resources is reduced. These data simplification methods can achieve the visualization of three-dimensional scenes when processing data of conventional scale.

[0003] As the scale of construction projects increases, the data density of the original model exceeds the processing capacity of general-purpose hardware. Existing solutions, when performing data compression, typically treat each component as a discrete geometric unit and assign simplification weights based on static business tags. Under this isolated simplification logic, the system ignores the physical support logic between components and the spatial assembly relationship, resulting in a loss of quality in the topological continuity of the simplified geometric model. This leads to high-value components losing accurate references in three-dimensional space. The transfer algorithm directly determines the reliability of the model engineering data. For example, Chinese invention patent CN119475520B discloses a lightweight transfer method for BIM data based on 3DE. It establishes hierarchical folders by identifying component tags, realizes the association between geometric information and attribute information, and ensures structural consistency. However, when performing mesh discretization processing, it lacks consideration of the mechanical transmission chain or assembly constraint path between components. Faced with high compression ratio requirements, the feature classification-based logic cannot ensure that the simplified connection interface coordinates remain locked, causing excessive assembly gaps between components.

[0004] Therefore, how to construct a non-uniform compression mechanism that takes into account both the semantic importance of components and the geometric topological continuity, so that the lightweight model can maintain rigorous engineering pose relationships while having fast loading performance, has become the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A lightweight conversion and high-performance real-time loading processing system for large-scale BIM data, comprising: The data parsing unit is used to read the original 3D model file and parse the data stream to extract the spatial coordinate dataset representing the geometric components, the business tag set representing the semantic attributes of the components, and the spatial topological relationship between the components. The spatial topological relationship includes the hierarchical dependency path of the components in the assembly tree. The association dependency analysis unit is used to determine the static attribute weight of the geometric component based on the semantic attributes of the component, and to identify the structural reference logic between each geometric component based on the spatial topological association relationship, so as to determine the topological constraint transmission factor that characterizes the degree of influence of spatial position constraints in the data chain. The comprehensive fidelity weight calculation unit is used to determine the comprehensive fidelity weight of a geometric component based on the linear superposition of the static attribute weight and the weights of adjacent components in a preset neighborhood space. The weight coefficient of the linear superposition value is mapped in real time by the topology constraint transfer factor, so that the component, which serves as the coordinate reference, obtains a high-level fidelity weight through the topology link. The topology feature locking unit is used to extract topology feature points that maintain the key features of the component outline in the spatial coordinate dataset based on the comprehensive fidelity weight, and associate the topology feature points with fidelity judgment parameters for constraining geometric deformation. The adaptive mesh processing unit is used to lock the absolute spatial coordinates of topological feature points when performing topological compression on the geometric mesh, and to perform mesh vertex merging according to the fidelity judgment parameters to generate a multi-level fineness model that maintains the consistency of spatial connection interfaces when changing multiple levels of fineness. Spatial retrieval organization units are used to build a lightweight retrieval framework that integrates global topological relationships and business semantic query paths, and to map multi-level fine-grained models as discretized spatial tiles that can be invoked on demand.

[0006] Preferably, the dependency analysis unit follows the following judgment rules when determining the topology constraint transfer factor: identify the connection dependency attribute between two adjacent geometric components; if the connection dependency attribute is a hard assembly constraint serving as a coordinate positioning reference, then the topology constraint transfer factor is set to a preset first transfer constant; if the connection dependency attribute is a non-positioning constraint serving only as spatial proximity, then the topology constraint transfer factor is set to a second transfer constant less than the first transfer constant, so that the low semantic component at the root node of the assembly tree can obtain a high level of geometric reference fidelity weight through the topology constraint transfer factor, thereby preserving the geometric features that serve as spatial positioning support in the mesh compression process.

[0007] Preferably, the spatial retrieval organization unit is used to construct a multidimensional relational semantic table containing geometric model data, time dimension data, and business management data, and to establish a streaming transmission protocol for discretized spatial tiles based on the lightweight retrieval framework. The streaming transmission protocol is used to prioritize pushing index data packets containing the overall spatial topology of the model when the client initiates a request, so that the client can reconstruct the spatial logical framework of the scene before the full geometric data arrives.

[0008] Preferably, the system also includes: a real-time parsing and rendering module, which, after the client receives the index data packet, performs frustum culling processing in conjunction with the client's hardware rendering capabilities, and initiates asynchronous data acquisition tasks based on the semantic path in the lightweight retrieval framework, thereby realizing progressive parsing and rendering of the model's geometric data.

[0009] Preferably, the system further includes: a loading priority calculation module, used to calculate the loading priority weight of each discretized spatial tile within the current viewport area based on user interaction trajectory analysis. The calculation formula is as follows: ,in, To load priority weights, The spatial distance between the geometric component and the current viewpoint, in units of . , The angle between the user's line-of-sight prediction vector and the component's position vector. and These are the preset distance weighting coefficient and direction weighting coefficient, respectively.

[0010] Preferably, the loading priority calculation module is used to initiate a data pre-request task for discretized spatial tiles within the prediction area in the background, and dynamically adjust the concurrent thread scale of the data pre-request task according to the real-time load of the client computing unit and the available network bandwidth.

[0011] Preferably, when generating multi-level fineness models, the adaptive mesh processing unit constrains the mesh collapse normal direction of topological feature points to limit the connection gap between components within a preset geometric tolerance range, so as to maintain the topological accuracy of the lightweight model in spatial verification scenarios.

[0012] Preferably, when generating discretized spatial tiles, the spatial retrieval organization unit adopts a hierarchical tree data structure, in which the top-level nodes store the outer envelope topological features of the model, the bottom-level nodes correspond to high-precision component geometric details, and the nodes at each level share the same topological feature points.

[0013] Preferably, the system further includes: a rendering frame rate monitoring unit, used to monitor the client's rendering frame rate in real time, the unit of which is... When the rendering frame rate is lower than the preset smoothing threshold, the dynamic replacement logic of discretized spatial tiles between different levels of fineness is triggered.

[0014] Preferably, the data parsing unit also includes a metadata cleaning subunit, which is used to perform normalized extraction of the original model data based on a preset industry standard semantic mapping library, and remove non-critical attribute fields that are unrelated to the spatial geometric benchmark.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. In lightweight conversion and high-performance real-time loading processing, the assembly constraints between components are extracted by constructing a topological correlation tensor matrix. This allows the fidelity requirements of high semantic level components to be transmitted to adjacent supporting or transitional components along the physical connection path, thereby correcting the geometric datum fidelity weights of low semantic level components. When performing triangular mesh simplification, the algorithm locks the vertex coordinates of the connection interface based on the weights, avoiding the component suspension or topological breakage caused by the simplification of business attributes in traditional solutions. This ensures that the lightweight model maintains rigorous engineering pose relationships under high compression ratios, giving the model the accuracy basis for spatial collaborative analysis such as support collision detection and clearance analysis.

[0016] 2. Relying on a hybrid index structure of spatial regions and professional systems, the system deconstructs large-scale model data into discrete logical units that are loaded on demand. With the help of a streaming protocol, the server prioritizes pushing a simplified index framework containing spatial topology, enabling the client rendering engine to rebuild the model framework and respond to the view frustum culling command before the full amount of geometric data arrives. This achieves deep parallelism between model loading and parsing rendering, eliminates the black screen waiting phenomenon in the traditional loading mode, and improves the initial response speed of large-scale models in lightweight environments such as mobile devices.

[0017] 3. By combining user interaction behavior trajectory analysis with dynamic matching of multi-level granularity models, the system silently initiates data preloading tasks for the prediction area in the background and ensures that models of different granularity levels inherit a unified physical connection surface as a shared topological anchor point across all levels during the generation phase. This ensures that the relative positional relationship between components remains logically consistent when the client performs dynamic switching of model levels, effectively eliminating geometric drift and visual jumps during model switching and achieving a balance between data loading performance and spatial logical accuracy. Attached Figure Description

[0018] Figure 1 This is a flowchart of the large-scale lightweight conversion and adaptive mesh generation of BIM data in this invention; Figure 2 This is a comparative analysis chart of the model rendering frame rate performance under different data compression rates in this invention. Figure 3 This is a diagram of the real-time loading system architecture of the cloud computing center and lightweight terminal in this invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] A lightweight conversion and high-performance real-time loading system for large-scale BIM data includes a data parsing unit, a dependency analysis unit, a comprehensive fidelity weight calculation unit, a topology feature locking unit, an adaptive mesh processing unit, and a spatial retrieval organization unit. The system extracts physical topological constraints between components and performs semantic weight propagation to achieve non-uniform simplification and dynamic scheduling loading of large-scale building information model data. The data parsing unit reads the original 3D model file and parses the data stream, extracting spatial coordinate datasets representing geometric components, business tag sets representing semantic attributes of components, and spatial topological relationships between components. These spatial topological relationships include the hierarchical dependency paths of components in the assembly tree. In large-scale engineering scenarios, to address the low parsing efficiency caused by the large number of components, the metadata cleaning subunit within the data parsing unit is based on preset industry standard semantics. The mapping library performs normalized extraction of the original model data and removes non-critical attribute fields unrelated to the spatial geometric datum. Through this processing, the system establishes a standardized geometric-attribute association graph, providing deterministic input data for subsequent weight calculations. When determining the execution logic of the metadata cleanup subunit, the system executes the business field attribute filtering procedure. The data parsing unit retrieves the industry-standard semantic mapping library stored in the local database, extracts the non-geometric attribute field names from the original 3D model file, and performs string weight matching with the spatial sensitive word set in the industry-standard semantic mapping library. If the field name belongs to a non-spatial association category such as construction progress, procurement batch, or contract number, it is determined as redundant metadata, and a jump read operation based on memory address offset is performed when parsing the data stream. By suppressing the memory usage of non-geometric metadata, the average parsing latency of a single component is controlled within a certain range. This provides a high-purity data source for the subsequent establishment of streaming protocols.

[0021] The dependency analysis unit determines the static attribute weights of the geometric component based on its semantic attributes. Based on spatial topological relationships, the structural reference logic between geometric components is identified to determine the topological constraint transfer factor, which characterizes the degree of influence of spatial position constraints in the data chain. The dependency analysis unit determines the topological constraint transfer factor. The following decision rules apply: Identify the connection dependency attributes between two adjacent geometric components; if the connection dependency attribute is a hard assembly constraint used as a coordinate positioning datum, then transfer the topology constraint transfer factor. Set to the preset first transfer constant; if the connectivity dependency attribute is only a non-location constraint based on spatial proximity, then set the topology constraint transfer factor. A second conduction constant, set to be less than the first conduction constant, allows low-semantic components at the root node of the assembly tree to pass through topological constraints. Obtain high-level geometric datum fidelity weights; the comprehensive fidelity weight calculation unit calculates the weights based on static attribute weights. The overall fidelity weight of a geometric component is determined by the linear superposition of the weights of its neighboring components within a predefined neighborhood space. Comprehensive fidelity weighting The calculation formula is as follows: ;in, The overall fidelity weight for geometric components, The static attribute weights of the geometric components. The weights of adjacent components within a preset neighborhood space. The topology constraint transfer factor is used; the weight coefficients of the linear superposition values ​​are determined by the topology constraint transfer factor. Real-time mapping allows components serving as coordinate references to obtain high-level fidelity weights via topology links, resolving the issue of high-semantic components being left unused due to oversimplification of low-semantic support components during the simplification process.

[0022] The dependency analysis unit determines the weights of static attributes. The process involves retrieving the built-in industry semantic weight database, comparing the parsed business tag set with preset classification fields, extracting the quantitative weights of each component, calculating the neighborhood superposition value, and constructing a detection system with the component's geometric center as the origin and a detection radius of [missing information]. The bounding box collision detection space is used to count the number of adjacent components that cause physical interference and the spatial hierarchy depth, according to the formula. The accumulated weights, where, For the comprehensive fidelity weight of geometric components, For static attribute weights, To pre-determine the weights of adjacent components within the neighborhood space, The topological constraint transfer factor. The average circumcircle diameter of scene components is set to 1.5 times to determine the topological propagation space range; the topological feature locking unit is based on comprehensive fidelity weights. The system extracts topological feature points that maintain the key features of the component outline from the spatial coordinate dataset and associates these feature points with fidelity judgment parameters used to constrain geometric deformation. The topological feature locking unit extracts the geometric anchor points of the component connection interfaces by calculating curvature gradients and identifying physical contact interfaces. The adaptive mesh processing unit locks the absolute spatial coordinates of the topological feature points when performing topological compression on the geometric mesh and performs mesh vertex merging according to the fidelity judgment parameters, generating a multi-level refinement model that maintains consistency in spatial connection interfaces during multi-level refinement transitions. When generating the multi-level refinement model, the adaptive mesh processing unit constrains the mesh collapse normal direction of the topological feature points, limiting the connection gaps between components within a preset geometric tolerance range, for example, by limiting them to... Within the specified range, the topological accuracy of the lightweight model is maintained in spatial verification scenarios. The adaptive mesh processing unit performs topological compression, employing an edge contraction algorithm based on a quadratic error metric matrix. The index positions of topological feature points in the spatial coordinate dataset are used as non-collapse anchor points as constraints. For non-feature point regions, a mesh vertex merging process is performed to calculate candidate edge contraction cost values, and a priority queue is established in ascending order. After each round of contraction operation, the mesh density at the component connection interfaces is monitored. With respect to preset geometric tolerances The proportional relationship is used to dynamically adjust the contraction step size. Local grid density, unit , A preset geometric tolerance of 0.001mm is used to ensure that multi-level refinement models share the same high-precision connection interface during dynamic switching, thus avoiding geometric drift caused by mesh simplification.

[0023] The spatial retrieval organization unit constructs a lightweight retrieval framework that integrates global topological relationships and business semantic query paths, and maps multi-level fine-grained models as discrete spatial tiles that can be invoked on demand. When generating discrete spatial tiles, the spatial retrieval organization unit adopts a hierarchical tree-like data structure, where top-level nodes store the model's outer envelope topological features, bottom-level nodes correspond to high-precision component geometric details, and nodes at each level share the same topological feature points. The spatial retrieval organization unit is used to construct a multi-dimensional relational semantic table containing geometric model data, time-dimensional data, and business management data, and establishes a streaming transmission protocol for discrete spatial tiles based on the lightweight retrieval framework. When a client initiates a request, the system prioritizes pushing an index data packet containing the overall spatial topology of the model, allowing the client to reconstruct the spatial logic framework of the scene before the full geometric data arrives. The loading priority calculation module calculates the loading priority weight of each discrete spatial tile within the current viewport area based on user interaction trajectory analysis. Loading priority weight The calculation formula is as follows: ;in, To load priority weights, The spatial distance between the geometric component and the current viewpoint, in units of , The angle between the user's line-of-sight prediction vector and the component's position vector. The preset distance weighting coefficient, The preset directional weight coefficients are used; the priority calculation module initiates a data pre-request task for the discretized spatial tiles within the prediction area in the background, and dynamically adjusts the concurrent thread scale of the data pre-request task according to the real-time load of the client computing unit and the available network bandwidth.

[0024] When performing viewpoint prediction based on user interaction trajectory analysis, the system executes a dead reckoning procedure based on temporal pose sampling. By collecting real-time translation, rotation, and zoom commands triggered by the user on the client, a local dynamic coordinate system with the current viewpoint coordinates as the origin is established. The magnitude and direction of the user's gaze prediction vector are determined based on the pose change rate within the previous three sampling periods. The priority calculation module then substitutes the user's gaze prediction vector as an input component into the formula. The operation, where the included angle The value of is corrected in real time through a linear compensation coefficient as the viewport rotation angular velocity increases, thereby ensuring that discretized spatial tiles on the predicted interaction path receive higher resource scheduling priority, and limiting the geometric data response latency during the 3D interaction process to within 100%. Within the visual delay threshold; load the priority calculation module to predict the user's gaze, in order to The sampling frequency acquires three consecutive periods of viewpoint pose parameters. The acceleration compensation component of the line-of-sight prediction vector is determined by calculating the second derivative of the viewpoint displacement vector, according to the formula... Output discretized spatial tile scheduling weights. To load priority weights, The spatial distance between the geometric component and the current viewpoint. The angle between the line-of-sight prediction vector and the component position vector, and the distance weighting coefficient. with direction weighting coefficient The ratio is dynamically linearly interpolated based on the available GPU memory on the client side. The loading time for high-priority tile data in the rendering sequence is within 50ms, ensuring the integrity of key components within the viewport during high-speed inspection of large-scale BIM data. The real-time parsing and rendering module, after receiving the index data packet on the client side, performs view frustum culling based on the client's hardware rendering capabilities and initiates asynchronous data acquisition tasks according to the semantic path in the lightweight retrieval framework, achieving progressive parsing and rendering of the model's geometric data. The rendering frame rate monitoring unit monitors the client's rendering frame rate in real time, in units of... When the rendering frame rate is lower than the preset smoothing threshold, the dynamic replacement logic of discretized spatial tiles between different levels of fineness is triggered. Through the collaboration of the above units, this invention achieves the improvement of the real-time loading performance of the model while maintaining high-precision topological constraints in large-scale BIM data application scenarios.

[0025] Example 1: In a system containing more than 20,000 process pipelines and their supporting facilities In the application scenario of building information modeling for underground integrated utility tunnels with support brackets, the geometric vertex scale of the raw data stream reaches [a certain number]. At level 1, mobile devices experience memory overflow when reading raw data streams. Conventional lightweighting methods classify support brackets as low-weight components based on component categories in the business tag set and assign them static attribute weights. When the topology transfer factor is set to 0.1, the spatial assembly relationship is ignored during mesh compression, causing displacement of the geometric feature points of the support frame. This results in the associated process pipelines appearing suspended in the 3D viewport, disrupting the spatial positioning reference in the computer-aided design data. When the system faces operational conditions, the dependency analysis unit identifies a rigid positioning constraint between the process pipelines and the support frame and transfers the topology constraint transfer factor accordingly. The first conduction constant is set at 0.8, and the comprehensive fidelity weight calculation unit receives static attribute weights from the process pipeline. It is 0.9, and according to the formula Calculate the overall fidelity weight of the support bracket. It is 0.82; among which, The overall fidelity weight for geometric components, The static attribute weights of the geometric components. The weights of adjacent components, The topology constraint transfer factor; the topology feature locking unit is based on the comprehensive fidelity weight. The topological feature points at the interface between the support bracket and the process pipeline are identified, and the absolute spatial coordinates of the topological feature points are locked by the adaptive mesh processing unit.

[0026] By transmitting the fidelity requirements of high semantic-level components along the topological constraint path to the supporting components, the contradiction between data volume reduction and maintaining physical assembly relationships is resolved. This ensures that the support bracket maintains consistency with the connection interface of the process pipeline during mesh collapse, limiting the connection gap between components to within 0.001mm. When the user's viewpoint movement speed exceeds 2000mm / s, the spatial retrieval organization unit predicts the angle between the user's line of sight vector and the component position vector. Calculate the loading priority weight of each discretized spatial tile. Loading priority weight The calculation formula is as follows: ;in, To load priority weights, The spatial distance between the geometric component and the current viewpoint, in units of , The angle between the user's line-of-sight prediction vector and the component's position vector. The preset distance weighting coefficient, The preset direction weight coefficients; the real-time parsing and rendering module assigns weights based on loading priority. By retrieving discretized spatial tiles and combining them with frustum culling, the mobile rendering frame rate is kept stable at 60Hz when the model's geometric data is compressed to less than 10% of the original volume. The assembly logic between components does not experience geometric drift as the level of the multi-level detail model changes, and the physical connection between the support bracket and the process pipeline remains intact during the 3D scene reconstruction process.

[0027] Example 2: In the verification test of real-time loading performance of ultra-large-scale computer-aided design models, the test platform simulated a mobile hardware environment with 4GB of memory and a network bandwidth limit of 10Mbps through a cluster computing system. The test data came from 10 sets of standard engineering files of building information models containing more than 500,000 geometric nodes. To verify the stability of the scheme under the uncertainty of actual measurement, Gaussian coordinate noise with a root mean square error of 0.5mm was injected into the original spatial coordinate dataset. The setting of the core parameter distance weight coefficient α depends on the ratio of the view frustum depth to the tile division granularity. It is used to adjust the balance between spatial retrieval accuracy and computing unit throughput load. When the ratio is at the lower limit of the value under high-frequency interaction, α tends to the upper limit of the value range to increase the weight of near-field components. Under the inspection condition, the distance weight coefficient α was set to 0.85 and the direction weight coefficient β was set to 0.15. The test process compared the sample group of the present invention and removed the topological constraint transfer factor. The first control group and the second control group, whose fidelity judgment parameters exceeded the preset range, were used to observe the assembly gap deviation between geometric components under different data compression rates. See Table 1. When the data compression rate increased from 50.0% to 95.0%, the first control group experienced positional shifts, with its average assembly deviation increasing from 1.251 mm to 15.422 mm, resulting in pipeline detachment from the support. This invention utilizes a comprehensive fidelity weighting method... The topological feature points were locked, and their average assembly deviation was maintained within 0.005mm. When the compression rate reached the performance inflection point of 98.0%, the second control group experienced a reduction in rendering frame rate to 18.2Hz due to excessive mesh collapse tolerance. The sample group of this invention maintained a rendering frame rate above 51.2Hz through the collaboration of adaptive mesh processing units and spatial retrieval organization units. Specific experimental data are as follows: Table #1: Comparison of test data for each group under different working conditions The above data confirms the topology constraint transfer factor. With comprehensive fidelity weight The introduction of this feature enables the system to maintain physical assembly logic while reducing the volume of geometric data. The adaptive mesh processing unit executes a coordinate locking procedure based on the fidelity judgment parameters, which solves the problem of the mutual constraint between semantic fidelity and topological continuity in the process of lightweighting computer-aided design data. It realizes the accuracy support of the model for collision checking and spatial verification tasks under extremely high compression ratios, and enables the model geometric data to achieve the goal of progressive analysis and real-time interaction in a resource-constrained environment.

[0028] Example 3: This example combines Figures 1 to 3 This section describes a lightweight conversion and high-performance real-time loading system for large-scale BIM data, such as... Figure 1 As shown, the original 3D model file is read, and the data parsing unit processes the input data to extract the component spatial coordinates, semantic attributes, and hierarchical paths. The data flows to the correlation and dependency analysis unit, which is responsible for identifying the reference logic to determine the topological constraint transfer factor. The processed data enters the comprehensive fidelity weight calculation unit, which determines the comprehensive weight based on the superposition of attribute weights and neighborhood weights. Then, the topological feature locking unit extracts feature points based on the calculation results and correlates them with fidelity judgment parameters. The adaptive mesh processing unit then performs topological compression and locks the feature point coordinates. The spatial retrieval organization unit constructs a retrieval framework and maps the model into spatial tiles, finally outputting discretized spatial tiles.

[0029] like Figure 2 As shown in the figure, the horizontal axis represents the data compression rate in %, with a scale covering the range of 50% to 98%, and the vertical axis represents the rendering frame rate in Hz, with a scale range of 0 to 65. The legend includes three sets of comparative data: the sample group of this invention, the first control group without topological constraints, and the second control group with parameter exceeding limits. Among them, the sample group of this invention maintains a high rendering frame rate throughout the entire compression range, the frame rate of the first control group decreases after the compression rate reaches 95%, and the second control group shows a trend of continuous decrease in frame rate as the compression rate increases; as shown in the figure. Figure 3 As shown, the high-performance computing center in the cloud on the left integrates modules for data parsing and dependency analysis, comprehensive fidelity weight calculation, topology feature locking and mesh compression, and spatial retrieval and organization of tiles. The generated discretized spatial tiles are stored in a discretized spatial tile library. The cloud connects to the lightweight interactive terminal on the right via a streaming protocol in an on-demand scheduling or index-first manner. The terminal contains a real-time parsing and rendering module for multi-level fine-grained model reconstruction, and an interaction priority calculation module for viewpoint perception and dynamic requests. The terminal sends request commands to the discretized spatial tile library to obtain the corresponding data according to the interaction requirements.

[0030] Example 4: In an industrial building auxiliary design scenario including a centrifugal pump set and a frequency converter control cabinet, the data parsing unit reads the original 3D model file and identifies the assembly constraint relationship between the pump housing and the anchor bolts. The pump housing has a high level of static attribute weight. Its value is 0.9, and the initial weight of the anchor bolt is 0.1; to solve the problem of excessive center distance of pump flange connection holes caused by global uniform scaling, the correlation dependency analysis unit executes the transmission constant calibration procedure based on the proportion of restricted connection degrees of freedom, identifies the physical contact properties of the connection interface of two adjacent geometric components, counts the number of restricted translational and rotational degrees of freedom, and calculates the formula Calculate the topology constraint transfer factor The specific calculation logic is as follows: the topological constraint transfer factor equals 1.0 minus the product of the total number of residual relative degrees of freedom and 0.03, where 0.03 is the weight attenuation coefficient per degree of freedom, obtained through load stiffness mapping experiments on 100 sets of underground utility tunnel support connection pairs. This represents that each unconstrained degree of freedom direction will cause a 3% decrease in topological constraint strength, thereby realizing the digital conversion from abstract connection logic to physical constraint strength. The topological constraint transfer factor. The total number of residual relative degrees of freedom between two geometric components, with dimensions of pure numbers; taking the assembly interface between the pump housing and the anchor bolts as an example, the system identifies that this interface contains fixed constraints restricting translation in three axes and rotation in three axes, and its total number of residual relative degrees of freedom is... The value is 0. Substituting this value into the formula yields the topology constraint transfer factor. The dependency analysis unit is set to 1.0, which represents the topological constraint transfer factor. The first transfer constant is set to 0.8; if the connection properties only have overlapping spatial bounding boxes without physical contact, then the topology constraint transfer factor is set to... The second conduction constant was determined to be 0.2, and the comprehensive fidelity weight calculation unit executed... The calculation transmits the high-fidelity requirements of the pump housing to the anchor bolts, increasing the overall fidelity weight of the anchor bolts. The value was adjusted to 0.82, thereby forcing the topology feature locking unit to lock the bolt center axis coordinates when processing spatial coordinate datasets.

[0031] When performing topology compression on the bolt mesh, the adaptive mesh processing unit maintains the absolute spatial position of the bolt mesh relative to the pump base plate based on the fidelity judgment parameters, thus limiting the assembly pose deviation of the bolt assembly to within a certain range. Within the tolerance range; when the system performs interoperation on a mobile device, the rendering frame rate monitoring unit collects the instantaneous frame rate of the past 10 rendering cycles in real time and calculates the sliding window average. Its unit is The system executes a negative feedback switching procedure: the rendering frame rate monitoring unit will adjust the sliding window average. The value is compared in real time with the preset 30Hz smoothing threshold; if the sliding window mean... For three consecutive sampling periods below 30Hz, the spatial retrieval organization unit retrieval lightweight retrieval framework loads priority weights within the current viewport. Discretized spatial tiles in the bottom 20.0% range are reduced by one level of refinement; if the sliding window mean... Once the refresh rate rises and stabilizes above 55Hz, the system triggers an incremental data recovery task for low-resolution tiles. The real-time parsing and rendering module then restores the details of the high-resolution model. By utilizing the consistency of topological feature points across different resolution levels, the system ensures that the relative spatial pose of the centrifugal pump casing and its associated piping interface does not experience geometric coordinate drift during resolution level changes. This enables complex industrial assembly models to maintain a 60Hz interactive rendering frame rate on mobile devices with only 4GB of memory, resolving the real-time parsing interruption issue caused by the mismatch between high-fidelity requirements and computational load.

[0032] Example 5: In a test scenario where the performance of terminal devices with different computing architectures is calibrated, the data parsing unit measures the number of instruction cycles and memory address seek time when the target hardware processes tiles with a preset geometric density, and establishes a hardware performance correlation table; the correlation dependency analysis unit determines the topology constraint transfer factor based on the hardware performance correlation table. The range of values ​​is determined by calculating the monotonically decreasing relationship between the pose deviation of the connection interface and the change of the conduction coefficient. The first conduction constant and the second conduction constant are determined to reduce the system bus load when the assembly tolerance of 0.001mm is met. The first conduction constant is 0.8 and the second conduction constant is 0.2. In this way, the weight parameters are constrained within the boundary of the terminal computing resources.

[0033] Under conditions of fluctuating graphics processing unit (GPU) kernel frequency, the rendering frame rate monitoring unit continuously collects the vertical synchronization wait time of 50 rendering frames during the 3D scene startup phase and calculates the standard deviation to determine a smoothing threshold reflecting the device's rendering stability. Its unit is ;in, For the smoothing threshold, The mean of the sliding window; spatial retrieval organizing units are based on a smoothing threshold. Set the sensitivity for adjusting the fineness level of the discretized spatial tiles, and monitor the mean of the sliding window during real-time interaction. With smoothing threshold The deviation component drives the loading priority weight. The allocation ratio offset allows the geometric data to be simplified in accordance with physical topological constraints, maintaining the pose of complex building components when changing multiple levels of refinement.

[0034] Example 6: In the case of performing offline tolerance calibration for a precision machine room process pipeline system, the data analysis unit measures the original curvature distribution of the standard flange connection and determines the set of feature normals at the component connection interface by executing a mesh boundary search algorithm; the adaptive mesh processing unit establishes an error cost function based on the set of feature normals to determine the maximum allowable collapse step size while meeting an absolute assembly accuracy of 0.001mm. ,in, ;in, The maximum collapse step size, in units of , For normal deviation tolerance, Local grid density, in units of By performing gradient descent iterations in the calibration procedure, the geometric anchor points of components with different semantic levels are locked when performing multi-level fineness replacements, thereby eliminating the topological overlap of component contact surfaces caused by step size settings, and stabilizing the relative pose deviation of the support and pipeline within the preset tolerance after simplification.

[0035] When faced with network bandwidth fluctuations below 100kbps, the spatial retrieval organization unit executes packet verification and reconstruction procedures, and calculates the data arrival delay of adjacent discretized spatial tiles by introducing timestamp sorting logic into the streaming protocol. Its unit is The priority calculation module is loaded according to the formula. Real-time adjustment of loading priority weights; among which, The corrected priority weights, To load priority weights, This is the time delay compensation coefficient, and its value is... , For data arrival delay; the system detects When the time exceeds 150ms, the downgrade processing of tile requests outside the viewport is triggered, and the rendering refresh of high-frequency flickering nodes within the current viewport is paused. This maintains the integrity of the spatial logic framework of the client rendering output and the stability of the scene structure during 3D interaction in the intermediate state where data has not fully arrived.

[0036] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0037] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A lightweight conversion and high-performance real-time loading system for large-scale BIM data, comprising: The data parsing unit is used to read the original 3D model file and parse the data stream to extract the spatial coordinate dataset representing the geometric components, the business tag set representing the semantic attributes of the components, and the spatial topological relationship between the components. The spatial topological relationship includes the hierarchical dependency path of the components in the assembly tree. The association dependency analysis unit is used to determine the static attribute weight of the geometric component based on the semantic attributes of the component, and to identify the structural reference logic between each geometric component based on the spatial topological association relationship, so as to determine the topological constraint transmission factor that characterizes the degree of influence of spatial position constraints in the data chain. The comprehensive fidelity weight calculation unit is used to determine the comprehensive fidelity weight of a geometric component based on the linear superposition of the static attribute weight and the weights of adjacent components in a preset neighborhood space. The weight coefficient of the linear superposition value is mapped in real time by the topology constraint transfer factor, so that the component, which serves as the coordinate reference, obtains a high-level fidelity weight through the topology link. The topology feature locking unit is used to extract topology feature points that maintain the key features of the component outline in the spatial coordinate dataset based on the comprehensive fidelity weight, and associate the topology feature points with fidelity judgment parameters for constraining geometric deformation. The adaptive mesh processing unit is used to lock the absolute spatial coordinates of topological feature points when performing topological compression on the geometric mesh, and to perform mesh vertex merging according to the fidelity judgment parameters to generate a multi-level fineness model that maintains the consistency of spatial connection interfaces when changing multiple levels of fineness. Spatial retrieval organization units are used to build a lightweight retrieval framework that integrates global topological relationships and business semantic query paths, and to map multi-level fine-grained models as discretized spatial tiles that can be invoked on demand.

2. The lightweight conversion and high-performance real-time loading processing system for large-scale BIM data according to claim 1, characterized in that, The dependency analysis unit follows these rules when determining the topology constraint transfer factor: it identifies the connection dependency attribute between two adjacent geometric components. If the connection dependency attribute is a hard assembly constraint serving as a coordinate positioning reference, the topology constraint transfer factor is set to a preset first transfer constant. If the connection dependency attribute is a non-positioning constraint serving only as spatial proximity, the topology constraint transfer factor is set to a second transfer constant less than the first transfer constant. This ensures that low-semantic components at the root node of the assembly tree obtain a high level of geometric reference fidelity weight through the topology constraint transfer factor, thereby preserving the geometric features that support spatial positioning during mesh compression.

3. The lightweight conversion and high-performance real-time loading processing system for large-scale BIM data according to claim 1, characterized in that, The spatial retrieval organization unit is used to construct a multidimensional semantic table containing geometric model data, time dimension data, and business management data. Based on the lightweight retrieval framework, a streaming protocol for discretized spatial tiles is established. The streaming protocol is used to prioritize pushing index data packets containing the overall spatial topology of the model when the client initiates a request, so that the client can reconstruct the spatial logical framework of the scene before the full geometric data arrives.

4. The lightweight conversion and high-performance real-time loading processing system for large-scale BIM data according to claim 1, characterized in that, The system also includes a real-time parsing and rendering module, which, after receiving the index data packet on the client, performs frustum culling in conjunction with the client's hardware rendering capabilities, and initiates asynchronous data acquisition tasks based on the semantic path in the lightweight retrieval framework to achieve progressive parsing and rendering of the model's geometric data.

5. The lightweight conversion and high-performance real-time loading processing system for large-scale BIM data according to claim 1, characterized in that, The system also includes a loading priority calculation module, used to calculate the loading priority weight of each discretized spatial tile within the current viewport area based on user interaction trajectory analysis. The calculation formula is as follows: ,in, To load priority weights, The spatial distance between the geometric component and the current viewpoint, in units of . , The angle between the user's line-of-sight prediction vector and the component's position vector. and These are the preset distance weighting coefficient and direction weighting coefficient, respectively.

6. The lightweight conversion and high-performance real-time loading processing system for large-scale BIM data according to claim 5, characterized in that, The loading priority calculation module is used to initiate data pre-request tasks for discretized spatial tiles within the prediction area in the background, and dynamically adjust the concurrent thread scale of the data pre-request tasks according to the real-time load of the client computing unit and the available network bandwidth.

7. The lightweight conversion and high-performance real-time loading processing system for large-scale BIM data according to claim 1, characterized in that, When generating multi-level refinement models, the adaptive mesh processing unit constrains the mesh collapse normal direction of topological feature points to limit the connection gap between components within a preset geometric tolerance range.

8. The lightweight conversion and high-performance real-time loading processing system for large-scale BIM data according to claim 1, characterized in that, When generating discretized spatial tiles, the spatial retrieval organization unit adopts a hierarchical tree data structure, in which the top-level nodes store the outer envelope topological features of the model, the bottom-level nodes correspond to high-precision component geometric details, and the nodes at each level share the same topological feature points.

9. The lightweight conversion and high-performance real-time loading processing system for large-scale BIM data according to claim 1, characterized in that, The system also includes a rendering frame rate monitoring unit, used to monitor the client's rendering frame rate in real time, with units of [unit missing]. When the rendering frame rate is lower than the preset smoothing threshold, the dynamic replacement logic of discretized spatial tiles between different levels of fineness is triggered.

10. A lightweight conversion and high-performance real-time loading processing system for large-scale BIM data according to claim 1, characterized in that, The data parsing unit also includes a metadata cleaning subunit, which is used to standardize and extract the original model data based on a preset industry standard semantic mapping library, and remove non-critical attribute fields that are not related to the spatial geometric benchmark.

Images