Cloud platform-based engineering quality detection data visual display system

Abstract

The application relates to the technical field of engineering quality detection data processing, cloud platform visualization and three-dimensional model rendering, in particular to an engineering quality detection data visualization display system based on a cloud platform, which is deployed on a cloud platform server and comprises the following modules: a data access module, which is used for acquiring multi-source heterogeneous detection data, historical evolution data and initial three-dimensional space model data carrying space coordinates, and completing cleaning, normalization and space mapping; an abnormality evaluation module, which is used for performing space-time feature analysis and dividing local abnormal data and global normal data; a visual routing module, which is used for generating a space visual gravity field and visual weight parameters; a topology reconstruction module, which is used for generating a local high-precision rendering model and a global degraded context model; and a dynamic rendering module, which is used for fusion display based on a space visual signal-to-noise ratio and dynamic adjustment of a context degradation compression ratio; the application realizes abnormality-driven visual autonomous reconstruction, and takes into account local risk identification and overall engineering context reservation.

Description

Technical Field

[0001] This invention relates to the fields of engineering quality inspection data processing, cloud platform visualization, and 3D model rendering technology, specifically to an engineering quality inspection data visualization system based on a cloud platform. Background Technology

[0002] Currently, the visualization of engineering quality inspection data is usually based on the unified overlay display of multi-source heterogeneous inspection results from building information models or cloud monitoring platforms. When the inspection data comes from multiple sources, has large differences in dimensions, and has both local anomalies and overall normal areas, the existing display methods cannot accurately focus on and render the abnormal areas in layers. Each display may include the synchronous loading and output of a large amount of low-priority background information, which reduces the accuracy of locating abnormal areas and the timeliness of terminal visualization rendering response. Summary of the Invention

[0003] The purpose of this invention is to provide a cloud-based engineering quality inspection data visualization system to solve the following technical problems:

[0004] By unifying massive amounts of multi-source engineering quality data with different dimensions into an interpretable spatial display chain, the display method is transformed from manual one-by-one anomaly retrieval to anomaly-driven visual autonomous reconstruction. While ensuring that the overall engineering spatial context is not lost, local areas that truly indicate quality risks are given priority in display resources, thus balancing the ability to identify local risks with the ability to preserve the overall engineering context.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A cloud-based engineering quality inspection data visualization system, deployed on a cloud platform server, includes:

[0007] The data access module is used to acquire multi-source heterogeneous detection data and corresponding historical evolution data carrying spatial coordinates from sensor gateways, detection agency data terminals and field inspection terminals through communication networks; acquire initial three-dimensional spatial model data containing multiple spatial nodes corresponding to the multi-source heterogeneous detection data; perform format cleaning and data normalization on the multi-source heterogeneous detection data; and map the normalized data to the corresponding spatial nodes of the initial three-dimensional spatial model data based on spatial coordinates.

[0008] The anomaly assessment module is used to perform spatiotemporal feature analysis on the mapped multi-source heterogeneous detection data, calculate the anomaly weight corresponding to each multi-source heterogeneous detection data, and divide the multi-source heterogeneous detection data into local abnormal data and global normal data based on the anomaly weight.

[0009] The visual routing module is used to calculate the distribution set of gravity values ​​of local abnormal data on each spatial node in the initial three-dimensional spatial model data based on the abnormal weight of local abnormal data, as a spatial visual gravity field, and generate visual weight parameters based on the spatial visual gravity field.

[0010] The topology reconstruction module is used to reconstruct the rendering topology of the initial 3D spatial model data according to the visual weight parameters, generate a local high-precision rendering model, and generate a global degraded context model according to the global normal data and the pre-configured initial context degradation compression ratio.

[0011] The dynamic rendering module is used to calculate the spatial visual signal-to-noise ratio of the local high-precision rendering model and the global degraded context model in terms of the percentage of pixels displayed on the screen. Based on the spatial visual signal-to-noise ratio, the local high-precision rendering model and the global degraded context model are sent to the client for fusion display. The visual weight routing response latency during the fusion display process is obtained, and the context degraded compression ratio is dynamically adjusted based on the visual weight routing response latency.

[0012] In one possible implementation, the specific process by which the anomaly assessment module calculates the anomaly weight corresponding to each multi-source heterogeneous detection data point includes:

[0013] Time series analysis is performed on the normalized multi-source heterogeneous detection data to extract time series values ​​containing timestamps;

[0014] Spatial analysis is performed on the normalized multi-source heterogeneous detection data to obtain spatial node data;

[0015] The difference between time series values ​​at adjacent time nodes is calculated as the mutation rate, and the absolute value of the difference between spatial node data and the preset spatial reference value is calculated as the threshold deviation.

[0016] Obtain the preset dimensionless spatiotemporal weight allocation coefficients with a sum of 1;

[0017] Based on the weighting coefficients, the mutation rate and threshold deviation are weighted and summed to obtain the anomaly weight corresponding to each multi-source heterogeneous detection data.

[0018] In one possible implementation, the specific process by which the anomaly assessment module divides multi-source heterogeneous detection data into locally anomalous data and globally normal data based on anomaly weights includes:

[0019] Obtain the preset anomaly detection threshold;

[0020] The anomaly weight corresponding to each multi-source heterogeneous detection data is compared with the anomaly judgment threshold;

[0021] If the anomaly weight is higher than or equal to the anomaly determination threshold, the corresponding multi-source heterogeneous detection data will be marked as local anomaly data.

[0022] If the anomaly weight is lower than the anomaly determination threshold, the corresponding multi-source heterogeneous detection data will be marked as global normal data.

[0023] In one possible implementation, the specific process by which the visual routing module calculates the spatial visual gravitational field of local anomaly data within the initial 3D spatial model data includes:

[0024] Obtain the spatial coordinates of local anomalies within the initial 3D spatial model data;

[0025] Calculate the spatial distance between each spatial node and local anomaly data in the initial three-dimensional spatial model data based on spatial coordinates;

[0026] Construct a gravitational decay function with anomaly weights as the numerator parameter and the exponential term of spatial distance as the denominator parameter;

[0027] The gravitational value of each spatial node in the initial three-dimensional spatial model data is calculated by the gravitational decay function, and the gravitational values ​​of each spatial node are combined to form a spatial visual gravitational field.

[0028] The maximum gravitational value in the spatial visual gravitational field is extracted as the gravitational extremum, and visual weight parameters are generated based on the gravitational extremum.

[0029] In one possible implementation, the topology reconstruction module performs rendering topology reconstruction on the initial 3D spatial model data based on visual weight parameters to generate a locally high-precision rendering model. The specific process includes:

[0030] Based on the visual weight parameters, determine the key rendering areas in the initial 3D spatial model data;

[0031] Extract the evolution feature vectors of historical evolution data within the key rendering area, as well as the geometric feature vectors of the initial 3D spatial model data of the key rendering area;

[0032] Based on the preset engineering risk feature mapping matrix, the evolutionary feature vector and the geometric feature vector are jointly projected onto the same risk semantic feature space for feature alignment; the cosine similarity between the aligned evolutionary feature vector and the geometric feature vector is calculated, and the cosine similarity is used as the fidelity parameter for multidimensional feature aggregation.

[0033] Based on multi-dimensional feature aggregation fidelity parameters, multi-level detail enhancement is performed on key rendering areas to generate local high-precision rendering models.

[0034] In one possible implementation, the specific process by which the topology reconstruction module generates a global degradation context model based on global normal data includes:

[0035] Obtain the remaining model area from the initial 3D spatial model data, excluding the key rendering area;

[0036] Map the global normal data to the remaining model region and calculate the initial rendering resource consumption of the remaining model region;

[0037] Get the current context downgrade compression ratio;

[0038] Based on the context-downgraded compression ratio, the remaining model region is processed to reduce texture resolution and suppress the number of rendered faces, generating a globally downgraded context model.

[0039] In one possible implementation, the specific process by which the dynamic rendering module calculates the spatial visual signal-to-noise ratio between the local high-precision rendering model and the global degraded context model includes:

[0040] Obtain the core pixel region corresponding to the local high-precision rendering model;

[0041] Obtain the background environment pixel region corresponding to the global degradation context model;

[0042] The first visual proportion of the number of pixels in the core pixel region to the total number of pixels on the screen is calculated, and the second visual proportion of the number of pixels in the background environment pixel region to the total number of pixels on the screen is calculated.

[0043] The quotient between the weight of the first visual perception and the weight of the second visual perception is calculated as the dynamic ratio, and the dynamic ratio is used as the spatial visual signal-to-noise ratio.

[0044] In one possible implementation, the dynamic rendering module fuses and displays the local high-precision rendering model and the global degraded context model based on the spatial visual signal-to-noise ratio, and dynamically adjusts the context degradation compression ratio. The specific process includes:

[0045] The local high-precision rendering model and the global degraded context model are merged and rendered in the same three-dimensional coordinate system, and the visual weight routing response latency during the fusion rendering process is calculated.

[0046] Obtain the preset delay threshold;

[0047] Compare the visual weighted routing response latency with a latency threshold;

[0048] If the visual weighted routing response latency is lower than or equal to the latency threshold, the current context degradation compression ratio is maintained, and the fused display result is output.

[0049] If the visual weighted routing response latency exceeds the latency threshold, the context degradation compression ratio is increased, and the topology reconstruction module is triggered to regenerate the global degradation context model for fusion display.

[0050] In one possible implementation, the multi-source heterogeneous detection data includes engineering quality stress-strain data, concrete strength data, and settlement displacement data;

[0051] The initial 3D spatial model data includes building information model component data.

[0052] The beneficial effects of this invention are:

[0053] 1) This invention cleans and normalizes the format of multi-source heterogeneous detection data carrying spatial coordinates and maps it to the corresponding nodes of a three-dimensional spatial model. This enables engineering data with different dimensions and sources to be directly compared under a unified discrimination framework, solving the problem of difficult collaborative analysis of multi-source heterogeneous data and laying the foundation for accurate anomaly identification.

[0054] 2) This invention performs spatiotemporal feature analysis on the mapped data, calculates the anomaly weight by combining the mutation rate of temporal changes and the deviation of spatial location, and accurately divides the data into local anomalies and global normality; this hierarchical mechanism effectively filters out the interference of slight fluctuations, prioritizes the use of limited display resources for core risk areas, and significantly improves the efficiency of engineering anomaly identification.

[0055] 3) This invention innovatively constructs a gravitational decay function based on anomaly weight and spatial distance to generate a spatial visual gravitational field; this mechanism transforms isolated anomaly judgment results into a visual attention distribution that decays smoothly from the risk core to the periphery, intuitively presenting the potential impact range of anomalies on surrounding adjacent components, which is highly consistent with the cognitive patterns of engineers conducting on-site investigations.

[0056] 4) This invention performs high-precision rendering of key areas with multiple levels of detail based on visual weight parameters, while downgrading the texture and face count of global normal data to generate a context model. This reconstruction method amplifies local risks and preserves evolutionary characteristics, while maintaining the overall structural context of the project under the condition that the system resource consumption is lower than the preset resource threshold, thus avoiding visual information overlap in the rendering interface.

[0057] 5) This invention optimizes the fusion display effect by calculating the spatial visual signal-to-noise ratio of the screen pixel ratio and monitors the visual weight routing response latency in real time; when the latency exceeds the limit, the system automatically and dynamically increases the context degradation compression ratio; this closed-loop mechanism effectively overcomes the stuttering caused by the synchronous loading of a large number of low-priority backgrounds, ensuring the timeliness and smoothness of high-precision visualization. Attached Figure Description

[0058] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0059] Figure 1This is a schematic diagram of a cloud-based engineering quality inspection data visualization system provided in an embodiment of this application. Detailed Implementation

[0060] 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 only some embodiments of the present invention, and 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.

[0061] Please see Figure 1 A cloud-based engineering quality inspection data visualization system, deployed on a cloud platform server, includes: a data access module, used to acquire multi-source heterogeneous inspection data and corresponding historical evolution data carrying spatial coordinates from sensor gateways, inspection agency data terminals and on-site inspection terminals through communication networks; acquire initial three-dimensional spatial model data containing multiple spatial nodes corresponding to the multi-source heterogeneous inspection data; perform format cleaning and data normalization on the multi-source heterogeneous inspection data; and map the normalized data to the corresponding spatial nodes of the initial three-dimensional spatial model data based on spatial coordinates;

[0062] The anomaly assessment module is used to perform spatiotemporal feature analysis on the mapped multi-source heterogeneous detection data, calculate the anomaly weight corresponding to each multi-source heterogeneous detection data, and divide the multi-source heterogeneous detection data into local abnormal data and global normal data based on the anomaly weight.

[0063] The visual routing module is used to calculate the distribution set of gravity values ​​of local abnormal data on each spatial node in the initial three-dimensional spatial model data based on the abnormal weight of local abnormal data, as a spatial visual gravity field, and generate visual weight parameters based on the spatial visual gravity field.

[0064] The topology reconstruction module is used to reconstruct the rendering topology of the initial 3D spatial model data according to the visual weight parameters, generate a local high-precision rendering model, and generate a global degraded context model according to the global normal data and the pre-configured initial context degradation compression ratio.

[0065] The dynamic rendering module is used to calculate the spatial visual signal-to-noise ratio of the local high-precision rendering model and the global degraded context model in terms of the percentage of pixels displayed on the screen. Based on the spatial visual signal-to-noise ratio, the local high-precision rendering model and the global degraded context model are sent to the client for fusion display. The visual weight routing response latency during the fusion display process is obtained, and the context degraded compression ratio is dynamically adjusted based on the visual weight routing response latency.

[0066] This embodiment provides a cloud-based visualization mechanism for engineering quality inspection data. Specifically, this embodiment takes a super high-rise complex project undergoing main structure acceptance as the main scenario. The project includes a basement floor slab, core tube shear wall, several transfer floor beam-column joints, and outer frame reinforced concrete components.

[0067] During construction, stress and strain monitoring, concrete strength testing, settlement and displacement observation, and corresponding historical inspection records are continuously connected. At the same time, the building information model updated synchronously with the project is stored in the cloud platform.

[0068] The goal of the system is not to display all multi-source heterogeneous detection data uniformly and without discrimination on the display terminal, but to give priority to display resources to local areas that are truly indicative of quality risks, while ensuring that the overall engineering spatial context is not lost.

[0069] The data access module collects detection data from different sources, including continuous time-series data from the sensor gateway, batch detection reports from third-party testing institutions, and point marking information from on-site inspection terminals;

[0070] Since various types of data are usually inconsistent in terms of units, sampling periods, field naming, and coordinate representation, format cleaning and data normalization are performed first.

[0071] The normalization here is not simply for computational convenience, but to enable engineering conditions with different dimensions, such as insufficient strength, increased settlement, and sudden strain rise, to be compared under the same discrimination framework. The system associates and maps the detection data to the corresponding spatial nodes in the three-dimensional model based on spatial coordinates, component numbers, or measurement point binding relationships. Taking the exemplary virtual verification model as an example, assuming there are nodes N1, N2, and N3 in the model, where N1 corresponds to the settlement monitoring point at the corner of the basement floor slab, N2 corresponds to the strain gauge at the beam end of the transfer layer, and N3 corresponds to the rebound detection area of ​​the core tube wall, the detection data from these three locations will be associated and mapped to the corresponding nodes respectively, and their historical evolution trajectory will be preserved.

[0072] After completing the spatial mapping, the anomaly assessment module performs spatiotemporal feature analysis on the data associated with each node; here, anomaly is not the same as a single violation of limits, but rather considers both the temporal trend and the degree of spatial deviation.

[0073] For example, if the rate of change of strain at the beam end exceeds the preset slope threshold per unit time, it usually means that there is a sudden change in the local stress path; while if the settlement data continues to deviate from other monitoring points in the same layer, it is more likely to reflect uneven deformation of the foundation.

[0074] Based on this, the system generates anomaly weights for each piece of detection data and distinguishes them into local anomaly data and global normal data. Local anomaly data represents areas that should be observed and interpreted first, while global normal data retains the overall background information of the project to maintain the observer's understanding of the structural location, adjacent components, and stress environment.

[0075] Based on this, the visual routing module transforms the anomaly weights into a spatial visual gravity field oriented towards the display layer; its technical meaning is: the closer a model node is to the anomaly region and the more it is affected by the anomaly state, the higher the display attention it should receive.

[0076] The further away a node is from the abnormal area and has no abnormal characteristics, the display precision is gradually reduced. This is in line with the cognitive pattern of engineers conducting on-site investigations, which is to first look at the core of the risk, then look at the adjacent transmission area, and then return to the global structural context.

[0077] After the visual routing module outputs the visual weight parameters, the topology reconstruction module reorganizes the rendering hierarchy of the 3D model accordingly, so that abnormal related components form a local high-precision rendering model, while other components without significant abnormalities form a global degraded context model.

[0078] Local models can improve mesh detail, enhance texture information, and overlay historical curves or status indicators; global models, on the other hand, can appropriately reduce the number of faces and texture resolution to save rendering computing resources in the cloud and on the terminal.

[0079] The dynamic rendering module calculates the spatial visual signal-to-noise ratio based on the visual proportion of the local and global models on the screen, and determines the fusion display method accordingly. The signal here corresponds to the abnormal core area that needs to be identified. The noise is not useless data, but refers to the background environment information that has low relevance to the current engineering safety status.

[0080] When the abnormal area is small but the risk is high, the system will increase the visual proportion of the area, the clarity of the outline, and the degree of aggregation of related information to avoid visual confusion caused by its local features in the global rendering; when the abnormal area spreads or the terminal bandwidth decreases, the system will dynamically adjust the context degradation compression ratio to ensure that the display remains responsive.

[0081] In terms of fault tolerance, if some detection data is missing spatial coordinates but has a clear component number or floor section identifier, the system can use the component primary key for substitution mapping.

[0082] If the timestamps of the same measurement point differ across multiple data sources, they are aligned according to a preset time window. The specific alignment process is as follows: a preset time window is generated with the target time node as the center, and the collected data from each data source within the window is uniformly resampled to the target time node using a linear interpolation algorithm.

[0083] Data that cannot be aligned is temporarily stored in a state pending verification and is not directly used for anomaly assessment; if a component in the initial 3D model has an inconsistent binding relationship with historical measurement points due to a version update, the system will first call the version mapping table to perform component inheritance matching.

[0084] If a match still cannot be found, the data will be displayed as a floating verification marker to avoid being mistakenly projected to the wrong location. If the cloud receives a large number of sudden anomalies in a short period of time, causing a single complete topology reconstruction to take too long, the system will first issue a lightweight anomaly selection result and highlight key components, and then complete a fine reconstruction in the background to ensure that the front end can see the risk location in time.

[0085] After the construction of the super high-rise complex entered the transfer layer, the cloud platform continuously received strain data from the beam-column joint at the southeast corner of the transfer layer, the rebound test results of the concrete strength in this area, and the observation data of the settlement points of the adjacent foundation.

[0086] After mapping these three types of data with the building information model components of the area, the system found that local nodes experienced accelerated stress changes in a short period of time, while adjacent settlement points also showed a continuous offset trend.

[0087] At this time, the system will not upgrade the entire building to the same high precision. Instead, it will automatically set the southeast corner transfer floor and its adjacent vertical force transmission path as the key observation range, display the beam-column nodes, slab edges and adjacent shear walls in a high precision model on the client, and at the same time, it will display other floors in a downgraded context.

[0088] Engineers can first see the abnormal focus area on the tablet terminal, and then view the historical evolution curves and the status of adjacent components superimposed on the same screen;

[0089] The purpose of this step is to integrate massive, multi-source, and different-scale engineering quality data into an interpretable spatial display chain, thereby transforming the display method from manual one-by-one anomaly retrieval to anomaly-driven visual autonomous reconstruction, while taking into account both the ability to identify local risks and the ability to preserve the overall engineering context.

[0090] In a preferred embodiment of the present invention, the specific process by which the anomaly assessment module calculates the anomaly weight corresponding to each multi-source heterogeneous detection data includes: performing time-series analysis on the normalized multi-source heterogeneous detection data to extract time-series values ​​containing timestamps; and performing spatial analysis on the normalized multi-source heterogeneous detection data to obtain spatial node data.

[0091] The difference between time-series values ​​at adjacent time nodes is calculated as the mutation rate, and the absolute value of the difference between spatial node data and the preset spatial benchmark value is calculated as the threshold deviation. A preset dimensionless spatiotemporal weight allocation coefficient with a sum of 1 is obtained. Based on the weight allocation coefficient, the mutation rate and the threshold deviation are weighted and summed to obtain the anomaly weight corresponding to each multi-source heterogeneous detection data.

[0092] This embodiment provides an abnormal weight generation mechanism; specifically, when judging based solely on the current value of a single measurement point, two types of misjudgments are likely to occur: one is that some data, although not yet reaching the absolute limit, deteriorate rapidly in a short period of time, and if only the current value is considered, they will be missed.

[0093] Secondly, some points may show little change in value, but their locations may deviate from the normal spatial distribution of similar structures. It is difficult to identify them in time if only the time curve is viewed. Therefore, this embodiment incorporates both time change characteristics and spatial deviation characteristics into the formation process of abnormal weights.

[0094] The system performs time-series analysis on each normalized detection data point and extracts time-stamped sequence values. For engineering detection scenarios, time-series characteristics reflect the drastic rate of component state evolution. For example, concrete strength usually increases steadily in stages and should not exhibit repeated jumps within a short period.

[0095] Structural settlement should evolve relatively slowly during normal consolidation. If the rate of change suddenly accelerates in a short period of time, it often means that the stress on the foundation or the groundwater disturbance conditions have changed. Accordingly, the system regards the degree of change between adjacent time points as the mutation rate to characterize the rate of deterioration.

[0096] The system also performs spatial analysis on the same data to extract its spatial nodes, floor sections, component categories and adjacent topological relationships; the spatial benchmark value can be set according to the engineering design value, the statistical stability interval of similar components, the group status of adjacent health measurement points or the benchmark monitoring surface confirmed by the supervisor;

[0097] Its engineering significance lies in characterizing the geometric deviation from the normal spatial reference position; for example, among multiple edge beams in the same transfer layer, even if the strain at one beam end has not yet exceeded the design alarm line, if it deviates significantly from the overall horizontal level of the adjacent beam ends on the same side, it may indicate uneven local force transmission; the system forms a threshold deviation based on this to characterize the degree of spatial anomaly.

[0098] In terms of weight allocation, the engineering significance of different data types is not entirely consistent; for settlement and displacement data, spatial deviation is usually more indicative of differential deformation problems; for stress and strain data, abrupt changes in time are more likely to indicate load redistribution or early signs of component damage.

[0099] For concrete strength data, it is more appropriate to consider the age, test batch, and stress location of the component. Therefore, the system introduces a preset weight allocation coefficient so that the mutation rate and threshold deviation can be combined according to engineering experience and project stage.

[0100] The specific combined calculation logic is as follows: multiply the mutation rate by the allocation coefficient for temporal variation, multiply the threshold deviation by the allocation coefficient for spatial deviation, and calculate the sum of the two as the anomaly weight of the multi-source heterogeneous detection data; this step provides a verifiable calculation process through weighted summation logic.

[0101] Taking the exemplary virtual verification model as an example, assume there are data D1 and D2. D1 corresponds to the strain at the beam end, which changes significantly in the last two sampling periods, while its spatial location is moderately different from that of similar beam ends. D2 corresponds to the settlement point, which changes gradually over time, but the height difference with adjacent points increases.

[0102] After different weighting coefficients are configured, D1 can reflect the priority of short-term risk, and D2 can reflect the priority of spatial non-uniformity, thereby avoiding data with different physical meanings being indiscriminately sorted on the same scale.

[0103] In terms of anomaly tolerance, if there are sampling gaps in the time series data, the system will only calculate the mutation characteristics when the minimum time continuity condition is met; if the condition is not met, the data will be marked as low confidence and will not be directly used as a high-priority anomaly source.

[0104] If the spatial reference value becomes invalid due to the change of construction conditions, such as before and after the support is removed, or before and after the transfer layer is poured, the system will automatically switch the corresponding reference template according to the construction stage; if the stage identification fails, it will fall back to the conservative reference confirmed by the supervisor to avoid misjudgment caused by reference drift.

[0105] If the same node is associated with multiple data sources but their conclusions conflict, the system retains the conflict state and does not force a merger. Instead, it prompts for multi-source objections during subsequent display to facilitate engineers' review.

[0106] In the aforementioned super high-rise project, strain gauges at the beam-column joint at the southeast corner of the transfer floor continuously uploaded data; the system found that the strain change rate of this joint in the most recent sampling periods was significantly higher than that in the previous construction stage, and its numerical distribution also gradually deviated from the overall range of adjacent beam ends on the same floor.

[0107] In contrast, although the current value of a certain settlement point in the basement has not risen sharply, the difference with the surrounding monitoring points is widening; the system combines the weight allocation strategy corresponding to the data type to form two abnormal weights from different sources but both with engineering orientation, providing a basis for subsequent local anomaly screening;

[0108] The purpose of this step is to incorporate both the trend of change and spatial deviation into the evaluation criteria, thereby achieving a more robust identification of early signs of instability, local differential deformation, and hidden stress anomalies.

[0109] In a preferred embodiment of the present invention, the specific process by which the anomaly assessment module divides multi-source heterogeneous detection data into local abnormal data and global normal data based on anomaly weights includes: obtaining a preset anomaly judgment threshold; and comparing the anomaly weight corresponding to each multi-source heterogeneous detection data with the anomaly judgment threshold.

[0110] If the anomaly weight is higher than or equal to the anomaly determination threshold, the corresponding multi-source heterogeneous detection data is marked as local anomaly data; if the anomaly weight is lower than the anomaly determination threshold, the corresponding multi-source heterogeneous detection data is marked as global normal data.

[0111] This embodiment provides a hierarchical screening mechanism for local anomalies and global normality. Specifically, if only anomaly weights are generated without explicit stratification, the system may still send a large number of slightly fluctuating points into the high-precision display queue, causing the truly critical risk areas to be obscured by visual congestion again.

[0112] Therefore, this embodiment sets up a hierarchical judgment process after the anomaly assessment, dividing the data into local abnormal data and global normal data, so that the limited display resources can be used to serve the engineering location that needs the most explanation.

[0113] The system has a preset anomaly detection threshold, which can be configured according to project type, construction stage, component importance and monitoring system; its engineering meaning is not simply that exceeding a certain value means danger, but rather serves as a trigger boundary for visual resource scheduling.

[0114] If the anomaly weight reaches or exceeds this boundary, it indicates that the data has sufficient spatiotemporal anomaly characteristics and should be included in the key observation area; if it is below this boundary, it is considered more suitable to be retained as background information of the overall project, and there is no need to trigger high-precision rendering separately.

[0115] Taking the exemplary virtual verification model as an example, if the strain data D1 associated with node N2 is determined to be a local anomaly, while the data corresponding to nodes N1 and N3 have not yet reached the triggering condition, the system will include the component where N2 is located and its adjacent force transmission area as key display candidates, while N1 and N3 continue to be part of the background context.

[0116] This classification aligns with actual engineering inspection practices; supervisors or design reviewers typically do not simultaneously examine all measurement points on a large screen or mobile device, but rather first identify which points are worth zooming in on immediately, and then understand the scope of risk in conjunction with their surrounding structure.

[0117] By dividing local anomalies into global normalities, the system can form a clear primary and secondary structure at the display layer, which avoids missing risk hotspots and also avoids mistaking minor disturbances as equally important events.

[0118] In terms of fault tolerance, if the abnormal weight of a certain data is close to the judgment threshold, the system can mark it as a transitional state and continue to observe it in one or more subsequent sampling periods.

[0119] If the boundary continues to hover, it can be included in the secondary attention list as needed for the project, but will not be immediately displayed at the highest precision.

[0120] If a project enters a special construction period, such as a phase that will cause short-term fluctuations, such as large-volume concrete pouring, formwork unloading, or support conversion, the system can enable phased threshold templates to prevent a large number of normal working condition fluctuations from being triggered erroneously. If the threshold configuration is missing or fails to be updated, the system will use the default conservative threshold and record the audit log simultaneously to avoid completely out-of-control abnormal generalization.

[0121] After the concrete pouring of the transfer layer in the super high-rise complex, the system calculated the abnormal weights of multiple beam-column joints and settlement points.

[0122] The strain data at the two beam ends in the southeast corner reached the key observation conditions and were marked as local abnormal data; most of the other nodes on the same floor only showed general construction disturbance, which was below the trigger boundary and was retained as global normal data.

[0123] In this way, the cloud platform will not elevate all components of the entire layer to the same display priority during the subsequent reconstruction phase, but will instead concentrate resources around the key nodes that have already been selected.

[0124] The purpose of this step is to establish a clear, hierarchical entry point for visual scheduling, thereby achieving a balance between focusing on anomaly information and preserving background information.

[0125] In a preferred embodiment of the present invention, the specific process of the visual routing module calculating the spatial visual gravitational field of local anomaly data in the initial three-dimensional spatial model data includes: obtaining the spatial coordinates of the local anomaly data in the initial three-dimensional spatial model data; and calculating the spatial distance between each spatial node in the initial three-dimensional spatial model data and the local anomaly data based on the spatial coordinates.

[0126] A gravity decay function is constructed with anomaly weights as the numerator parameter and the exponential term of spatial distance as the denominator parameter. The gravity value of each spatial node in the initial three-dimensional spatial model data is calculated through the gravity decay function, and the gravity values ​​of each spatial node are combined to form a spatial visual gravity field. The maximum gravity value in the spatial visual gravity field is extracted as the gravity extremum, and visual weight parameters are generated based on the gravity extremum.

[0127] This embodiment provides a spatial visual gravity field generation mechanism; specifically, after completing the local anomaly screening, if the system simply highlights the anomaly points individually, although it can show where the problem is, it is difficult to express the potential impact range of the anomaly on the surrounding components, and it is also not conducive to building a continuous visual transition.

[0128] Especially in engineering structures, risks often spread along the stress path, connection nodes, and adjacent components. Displaying only a single point would distort the semantics of the project. Therefore, this embodiment introduces a spatial visual gravity field to expand the impact of anomalous data from a single point to a spatially significant distribution with gradients.

[0129] The system acquires the spatial coordinates of local anomaly data in the 3D model; these coordinates can be the location of a monitoring point, or the center of a component, a surface area of ​​a component, or a bounding box of a component that is bound to a certain detection result; the system calculates the spatial distance from other spatial nodes in the model to the location of the anomaly;

[0130] The distance here not only serves as a geometric measure, but also reflects the natural judgment habits of engineers when observing nearby affected areas: the closer the node is to the risk source, the more it usually needs to be checked together; the farther away the area is from the risk source, unless there are other anomalies, it does not need to occupy too much visual resources.

[0131] When a gravitational field is formed, the anomaly weight acts as a source of risk intensity, while the spatial distance plays a role in the attenuation of the influence; that is to say, the stronger the anomaly, the larger the visual attention range.

[0132] The greater the distance, the lower the attention level becomes. It should be noted that, to avoid algorithmic errors caused by zero denominator and infinite gravity due to zero computational distance between local outlier data and their corresponding spatial nodes, a preset non-zero constant is added to the denominator parameter as a smoothing factor. The system establishes the gravitational decay function as:

[0133]

[0134] in, This represents the calculated gravitational value of a spatial node. This indicates the anomaly weight of the local outlier data. This indicates the spatial distance between the spatial node and the abnormal data. The dimensionless attenuation index parameter is calibrated based on the spatial propagation laws of historical engineering data. A preset system tolerance constant is used to prevent system overflow caused by the denominator approaching zero.

[0135] Through this attenuation relationship, the system can form a visual hierarchy that smoothly transitions from the core of the risk to the periphery, rather than a rigid on / off switch.

[0136] Taking the exemplary virtual verification model as an example, if the anomaly is located at node N2, then the nodes N2-1 and N2-2 directly connected to it will obtain a high gravity value, while nodes in the same layer but with a spatial distance greater than a preset distance threshold will only obtain a medium to low gravity value, and nodes across layers and without a direct structural connection will have a very low gravity. When multiple anomaly sources exist at the same time, their respective gravity distributions can be superimposed, ultimately forming a visual attention hotspot in the entire engineering space.

[0137] The system extracts the gravitational extremum from the gravitational field and generates visual weight parameters accordingly. Specifically, the system divides the gravitational values ​​of each node in the spatial visual gravitational field by the gravitational extremum for global normalization, and converts the result into a relative proportion parameter distributed in the interval of 0 to 1, which is used as the visual weight parameter of the node.

[0138] The gravity extremum represents the core position that needs to be highlighted in the current scene. The visual weight parameter can further determine the focus of the viewpoint, the range of rendering level improvement, the degree of color contrast enhancement, and the priority order of unfolding related curves or explanatory cards. In this way, the display logic no longer depends on manual selection, but is automatically driven by the spatial risk distribution.

[0139] In terms of fault tolerance, if a certain abnormal data lacks precise coordinates but the component can be located, the system uses the center point of the component's bounding box or the center of the key surface as an approximate location.

[0140] If multiple abnormal points overlap densely, causing excessive concentration of local gravitational field, the system can process them by component family or floor cluster to avoid local display pixel overlap exceeding the preset threshold.

[0141] If the distance calculation results show interpretation deviations caused by cross-component occlusion or floor partitions, the system can correct the propagation range by combining the topological relationship of the building information model. For example, it can prioritize propagating attention along connecting components rather than simply penetrating unrelated spaces.

[0142] If an abnormal weight is higher than the preset sensor fault analysis threshold but is determined to be a sensor fault after manual review, the system can remove it from the gravity source list, retain only the fault label and not trigger visual focusing;

[0143] In the aforementioned project, after the beam-column joint of the southeast corner transfer layer was identified as a local anomaly, the system used this joint as the spatial source point to generate a visual attention distribution with progressively decreasing intensity to adjacent beam segments, column base areas, edges of the same-layer slab, and upper load-bearing components.

[0144] What engineers see on the client is not an isolated, flickering single measuring point, but a key observation area that extends from the abnormal node to the adjacent components. The beam-column interface is the clearest, the surrounding components are the next clearest, and other floors retain an outline background.

[0145] The purpose of this step is to transform discrete anomaly determination results into a continuous spatial attention structure, thereby achieving risk range expression and targeted delivery of visual resources in accordance with engineering cognitive logic.

[0146] In a preferred embodiment of the present invention, the specific process of the topology reconstruction module performing rendering topology reconstruction on the initial three-dimensional spatial model data according to the visual weight parameters to generate a local high-precision rendering model includes: determining the key rendering region in the initial three-dimensional spatial model data according to the visual weight parameters; extracting the evolution feature vector of the historical evolution data in the key rendering region, and the geometric feature vector of the initial three-dimensional spatial model data in the key rendering region;

[0147] Based on the preset engineering risk feature mapping matrix, the evolutionary feature vector and the geometric feature vector are jointly projected onto the same risk semantic feature space for feature alignment; the cosine similarity between the aligned evolutionary feature vector and the geometric feature vector is calculated, and the cosine similarity is used as the multi-dimensional feature aggregation fidelity parameter; based on the multi-dimensional feature aggregation fidelity parameter, the key rendering areas are enhanced with multiple levels of detail to generate a local high-precision rendering model.

[0148] This embodiment provides a local high-precision rendering model generation mechanism; specifically, spatial visual gravitational field alone is not enough to ensure that key areas are correctly magnified.

[0149] If high-precision improvement is based solely on the current location of the anomaly without considering the historical evolution and geometry of the area, two problems can easily arise: First, the anomaly information that should have spread along cracks, joints, or stress boundaries is incorrectly and uniformly mapped onto the entire surface of the component.

[0150] Secondly, when zooming in locally, key details related to defect evolution are lost, making the display look clear but lacking in explanatory power. Therefore, after determining the key rendering areas, this embodiment constructs and introduces multi-dimensional feature aggregation fidelity to guide the improvement of detail levels.

[0151] The system determines the key rendering area based on visual weight parameters; this area can be a single component, a local sub-model composed of several adjacent components, or a local surface patch on a component that corresponds to the detection result; the system extracts the evolution feature vector of the historical evolution data within the key rendering area, as well as the geometric feature vector of the initial three-dimensional spatial model data of the area;

[0152] The evolutionary feature vector here reflects how the risk develops over time, such as whether the settlement is a continuous accumulation, or whether the strain is a sudden increase or a periodic oscillation; the geometric feature vector reflects the shape, boundaries, connection relationships and weak points of the component, such as beam-column joints, opening edges, and abrupt changes in cross-section, which are usually more likely to become areas of abnormal concentration.

[0153] It is important to emphasize that, since evolutionary feature vectors belong to the time series dimension and geometric feature vectors belong to the spatial morphology dimension, there is an inherent dimensional misalignment and semantic heterogeneity between the two. Therefore, it is mathematically impossible to directly calculate the cosine similarity.

[0154] Therefore, the system has a built-in shared engineering risk feature mapping matrix. First, through feature alignment operation, the temporal evolution feature sequence and spatial geometric feature parameters are jointly projected into the risk semantic feature space of the same dimension, so that they are comparable in the same mathematical space, and then the cosine similarity calculation is performed.

[0155] The specific construction method of the engineering risk feature mapping matrix is ​​as follows: extract the temporal evolution feature samples and spatial geometric feature samples of known risk samples in historical projects, and construct a dimension-reduced projection matrix that maps the original temporal and spatial feature dimensions to the same preset feature dimension space based on principal component analysis or autoencoder network; perform matrix multiplication operations with the evolution feature vector and geometric feature vector respectively through the projection matrix to complete the feature alignment operation.

[0156] The specific structured calculation steps are as follows: calculate the ratio of the inner product of the two vectors after projection alignment to the product of their magnitudes, and use this ratio as the cosine similarity; perform correlation calculation on these two types of unified features. The purpose is not to make an abstract mathematical demonstration, but to determine whether the historical change pattern is consistent with the engineering sensitive form of the geometric region.

[0157] When the historical evolution trend is highly consistent with the geometrically sensitive area, it indicates that the local anomaly is more likely caused by the actual stress on the component or changes in the material state. At this time, the system adopts a higher level of detail enhancement in this area, such as retaining a finer mesh structure, enhancing crack texture expression, overlaying more historical state slices, or displaying denser measurement point annotations.

[0158] Conversely, if the correlation between the two is weak, it indicates that the anomaly may be more likely to be an isolated measurement point anomaly, external disturbance, or a state that is yet to be verified. The system can still improve the display, but will control the improvement range to avoid excessive amplification.

[0159] Taking the exemplary virtual verification model as an example, if the evolution characteristics of the key region R1 match its geometric boundary, it indicates that the anomaly is spreading along the beam-column node transmission path, and the system adopts a higher level of detail for R1; if region R2 has local numerical anomalies, but its historical characteristics are inconsistent with its geometric weaknesses, the system only performs limited enhancements in the neighborhood of the measurement point.

[0160] In terms of anomaly tolerance, if the historical evolution data of a key area is insufficient, for example, the newly deployed measuring points have not yet formed a complete time series, the system will revert to the high-precision reconstruction strategy based on the current anomaly weight and geometrically sensitive location, and add a prompt that the area has insufficient historical samples.

[0161] If the geometric model is too simplified and cannot provide reliable boundary features, the system can call higher version building information model components, as-built scan models, or component family standard templates to supplement geometric details;

[0162] If it is still not possible to fill the gap, then excessive refinement should be restricted to avoid visual details deviating from the actual structure. If the historical evolution and geometric features are completely in conflict, for example, if the data indicates that the anomaly is located in the center of the component while the component's historical defects are mainly concentrated in the edge connection area, then the system should retain the conflict display and prompt manual review, rather than forcing convergence to a single conclusion.

[0163] In this super high-rise project, the system identified the area around the beam-column joint of the southeast corner transfer layer as the key rendering area. Through comparison of the feature mapping matrix, it was found that the historical changes of strain anomalies in this area were highly consistent with the geometrically stress-sensitive area at the junction of the beam end and column edge. At the same time, the retest results of the nearby concrete strength also showed that there was a relatively weak section in this junction zone.

[0164] Therefore, the system enhances the mesh details in this local area, improves the surface texture display of the node core area, and overlays the historical evolution trajectory of several detections in this area, so that engineers can see the construction relationship and understand how the anomaly gradually develops in the same view;

[0165] The purpose of this step is to maintain the correspondence between abnormal information and structural morphology during local magnification, thereby achieving high-precision display with realism, interpretability, and engineering relevance.

[0166] In a preferred embodiment of the present invention, the specific process of the topology reconstruction module generating a global degradation context model based on global normal data includes: obtaining the remaining model regions in the initial three-dimensional spatial model data excluding the key rendering regions; mapping the global normal data to the remaining model regions; and calculating the initial rendering resource consumption of the remaining model regions.

[0167] Obtain the current context degradation compression ratio; based on the context degradation compression ratio, perform texture resolution reduction and rendering polygon count suppression on the remaining model regions to generate a global degradation context model.

[0168] This embodiment provides a global degradation context model generation mechanism; specifically, a local high-precision model can highlight the core of the anomaly, but if the rest of the area is completely hidden, engineers often lose the basis for judging the floor where the anomaly is located, adjacent components, structural boundaries and the overall stress environment.

[0169] Conversely, if the remaining areas are still preserved at full precision, mobile terminals and web pages are prone to noticeable lag. Therefore, this embodiment introduces a global degradation context model in addition to local high precision to preserve the overall picture of the project at a lower resource cost.

[0170] The system extracts the key rendering areas from the initial 3D spatial model to obtain the remaining model areas; then it maps the data that was previously determined to be globally normal into these areas, so that they continue to serve as the expression of structural background and normal operation.

[0171] The system assesses the initial rendering resource consumption of the remaining area, including the number of model faces, material complexity, texture size, and additional annotation density; based on the current context degradation compression ratio, it performs texture resolution reduction, rendering face count suppression, annotation density convergence, or long-distance component merging on the remaining area.

[0172] Its global degradation process does not involve indiscriminate pixel-level blurring of the background, but rather compresses visual details that are less relevant to the current anomaly detection without destroying key topological relationships.

[0173] Context degradation compression ratio can be understood as the degree to which background information relinquishes resources; the more concentrated the anomalies and the more limited the terminal's computing power, the higher the compression ratio can be; when multiple adjacent areas need to be compared simultaneously or the terminal's display capabilities are strong, the compression ratio can be relatively reduced to retain more context details.

[0174] Taking the exemplary virtual verification model as an example, if the key area is located in the southeast corner of the 20th floor, the remaining components of the 20th floor, the adjacent force transmission areas of the 19th and 21st floors can be slightly downgraded, while the high-rise outer frame components far from this area and the large area of ​​the basement can be downgraded more strongly, but the floor number, component outline and main passage relationship are still retained.

[0175] In terms of fault tolerance, if there are objects in the remaining area that are critical lifeline components, such as the core tube main wall, transfer truss, and main support columns, even though they have not reached the fault threshold, the system can set a minimum display fidelity baseline for them and will not allow excessive simplification due to background degradation.

[0176] If the terminal device is a high-performance workstation, the compression ratio can be appropriately reduced to enhance the integrity of the global context; if the terminal is a mobile inspection terminal and the network is limited, priority should be given to compressing remote non-critical areas to ensure that local high-precision areas are smoothly visible.

[0177] If abnormal data is subsequently received in a certain background area, that area will immediately be removed from the normal degradation queue and re-participate in the key rendering area selection to avoid background information lag.

[0178] During the abnormal display process of this super high-rise project, the system treated the southeast corner transfer layer and its adjacent force transmission components as local high-precision models, while treating the remaining floors, the outer frame facade and most of the underground structure as context areas.

[0179] For the background, the system retains the floor outline, main beam column lines and component numbering frame, but reduces the display of detailed textures and high-density measuring point markers;

[0180] In this way, engineers can still see which side of the building the anomaly is located on, which structural transition zone it belongs to, and how far it is from the core tube, without causing view lag due to detailed rendering of the entire building;

[0181] The purpose of this step is to preserve the overall context of the project while saving rendering resources, thereby achieving a display effect that highlights key points without distortion of the overall picture.

[0182] In a preferred embodiment of the present invention, the specific process of the dynamic rendering module calculating the spatial visual signal-to-noise ratio between the local high-precision rendering model and the global degraded context model includes: obtaining the core pixel region corresponding to the local high-precision rendering model; and obtaining the background environment pixel region corresponding to the global degraded context model.

[0183] The first visual proportion of the number of pixels in the core pixel region to the total number of pixels on the screen is calculated, and the second visual proportion of the number of pixels in the background environment pixel region to the total number of pixels on the screen is calculated. The quotient between the first visual proportion and the second visual proportion is calculated as the dynamic ratio, and the dynamic ratio is used as the spatial visual signal-to-noise ratio.

[0184] This embodiment provides a spatial visual signal-to-noise ratio calculation mechanism; specifically, even if a local high-precision model and a global downgraded model have been formed, whether the final display effect is truly beneficial to engineering judgment still depends on the visual relationship between the two on the screen.

[0185] If the percentage of pixels in the abnormal area displayed on the screen is lower than the first preset percentage threshold and it is surrounded by a background area with more than the preset upper limit of pixels, the observer may still have difficulty locating the risk in time; if the percentage of pixels in the abnormal area displayed on the screen is higher than the second preset percentage threshold, it may lose the reference of its surrounding structural environment; therefore, this embodiment uses the spatial visual signal-to-noise ratio to measure whether the key information is prominent enough and does not excessively deviate from the background.

[0186] The system obtains the core pixel region of the local high-precision rendering model in the current view, and the background environment pixel region of the global degradation context model in the current view; the core pixel region represents the visible projection of the risk signal, and the background environment pixel region represents the visible projection of the global engineering context.

[0187] The proportion of these two on the screen essentially corresponds to the allocation of an engineer's attention to risks and the environment in the current observation interface; if the proportion of the core pixel area is too low, it means that the key area is too small in the view, and abnormal information is difficult to form a stable visual recognition.

[0188] If the background pixel area accounts for too low a proportion, it means that the context is compressed too much, and it may be difficult to determine the relationship between the anomaly and adjacent components, floors and structural boundaries.

[0189] By calculating the dynamic ratio between the two, the system can quantify whether the current view is suitable as a key inspection screen. When performing the division calculation, in order to prevent the user from triggering a program crash error due to the background environment pixels being completely invisible and the second visual weight dropping to zero under an extreme local magnification view, the system presets a non-zero safety minimum lower limit for the second visual weight. When the calculated second visual weight is lower than this lower limit, the lower limit value is forcibly used for the division calculation.

[0190] The spatial visual signal-to-noise ratio here is not the same as the concept of noise in image processing, but is used to describe the relative perceptibility of abnormal core information in the current engineering interface relative to the background environment information;

[0191] Taking the exemplary virtual verification model as an example, if the core area occupies only a small part of the screen in the current view, while the background occupies most of the screen, the spatial visual signal-to-noise ratio is low, and the system should tend to further focus on the core area or compress the background details.

[0192] If the core area already occupies a large proportion and the background can still provide sufficient information on layer and component positioning, then the current display is more suitable for quick assessment.

[0193] In terms of fault tolerance, if the user actively zooms or rotates the view, causing the core area to temporarily move off the edge of the screen, the system will re-identify the core pixel area and provide a regression view suggestion when necessary to avoid distortion of the signal-to-noise ratio calculation.

[0194] If multiple scattered abnormal regions enter the screen at the same time, the system can identify multiple core pixel clusters separately and comprehensively evaluate the current signal-to-noise structure based on the total core proportion and dispersion.

[0195] If the background environment is significantly reduced due to occlusion or sectioning of the view, the system adjusts the signal-to-noise ratio threshold in combination with the view mode to avoid incorrectly judging the background as insufficient in the section inspection scenario.

[0196] During the aforementioned abnormality display, the engineer viewed the anomaly of the transition layer from the overall building overview perspective; at this time, the key area in the southeast corner only occupied a small part of the image, while the background floors and facade outline occupied most of the screen. The system recognized that the visual proportion of the core area was low, so it automatically suggested switching to a local view centered on the southeast corner transition layer.

[0197] When the perspective is adjusted, the proportion of key areas increases, while the core tube, adjacent beams and columns and floor boundaries remain in the background. The system determines that the new spatial visual signal-noise structure is more suitable for anomaly interpretation.

[0198] The purpose of this step is to evaluate the display effect using quantifiable screen ratios, thereby achieving a dynamic balance between view focus and context retention.

[0199] In a preferred embodiment of the present invention, the dynamic rendering module fuses and displays a local high-precision rendering model and a global degraded context model based on the spatial visual signal-to-noise ratio, and dynamically adjusts the context degradation compression ratio. The specific process includes: performing scene fusion rendering of the local high-precision rendering model and the global degraded context model in the same three-dimensional coordinate system, and calculating the visual weight routing response delay during the fusion rendering process; obtaining a preset delay threshold.

[0200] The visual weighted routing response delay is compared with the delay threshold. If the visual weighted routing response delay is lower than or equal to the delay threshold, the current context degradation compression ratio is maintained and the fusion display result is output. If the visual weighted routing response delay is higher than the delay threshold, the context degradation compression ratio is increased and the topology reconstruction module is triggered to regenerate the global degradation context model for fusion display.

[0201] This embodiment provides a mechanism for integrating display and dynamic compression ratio adjustment; specifically, the aforementioned mechanism has solved the problems of where to look and how clearly to look, but in practical applications on engineering sites, the problem of whether it can be seen in a timely manner must also be solved.

[0202] If every abnormal trigger is accompanied by a large-scale re-rendering and the response delay exceeds the acceptable range for engineers, then even if the system displays reasonable logic, it will be difficult to use for fast-paced decision-making on site.

[0203] Therefore, this embodiment constructs a visual weighted routing response delay to perform closed-loop control, and dynamically adjusts the context degradation compression ratio according to the actual response situation;

[0204] The system merges the local high-precision rendering model with the global degraded context model in the same three-dimensional coordinate system to ensure that key areas and background structures are strictly aligned in space; in this way, engineers will not lose the accurate position of local details in the whole building when viewing them.

[0205] During the integration process, the system monitors the response latency from anomaly triggering, visual weight updating, model reconstruction to the client completing the visualization switch; the engineering significance of this latency is to measure whether the system can keep up with changes in on-site quality status and the pace of manual inspection;

[0206] The system calculates a preset latency threshold based on the hardware rendering refresh rate of the terminal device and the channel transmission bandwidth of the current communication network. The specific setting logic is as follows: extract the minimum screen refresh cycle expected by the terminal device. And obtain the average round-trip time of network data packets. Multiply the sum of the two by the preset safety margin constant. As a delay threshold;

[0207] If the visual weight routing response latency of the current view is lower than or equal to the latency threshold, the current context degradation compression ratio is maintained, indicating that the system can still complete the visual reconstruction in a timely manner under the existing background fidelity.

[0208] If the response delay exceeds the threshold, it indicates that the background model is consuming too many resources or the current terminal / network environment is insufficient. In this case, the system increases the context degradation compression ratio, regenerates a lighter background model, and merges it with the local high-precision model again.

[0209] Its core idea is that the visibility of the abnormal core area takes precedence over the integrity of background details. As long as the basic engineering context is not destroyed, the background should give way to key areas and overall response speed.

[0210] Taking the exemplary virtual verification model as an example, assuming that the current client has already loaded the key area R1 and the background area C1, when a new adjacent anomaly source is added, the system needs to reorganize the view;

[0211] If the background still retains too many textures and patches during the first fusion, resulting in a slow response, the system will adjust C1 to C2 with a higher degree of compression, retaining only the floor outline, key component outlines, and necessary annotations, before fusion and display; in this way, the high-precision display of R1 and the overall refresh rate can be guaranteed at the same time.

[0212] In terms of fault tolerance, if the abnormal response delay is not caused by rendering resources, but by network jitter, insufficient client memory or model version conflict, the system can prioritize triggering link diagnosis instead of simply increasing the compression ratio to prevent ineffective degradation.

[0213] If the compression ratio has been increased to the preset limit but the latency still exceeds the limit, the system will switch to emergency display mode, and only the high-precision sub-model of the key area and the minimum skeleton background will be sent first. The context will be completed after the link is restored.

[0214] If the user manually locks certain background components that should not be downgraded in the background, such as the core tube or key support system, the system will retain the baseline accuracy of these objects when recalculating the compression ratio, and only compress other non-critical background areas.

[0215] In this super high-rise project, after the anomaly of the southeast corner transfer layer was triggered, the system distributed the local high-precision model and the degradation context model of the entire building to the tablet terminals of the project general contractor and supervisor. During the initial demonstration, the on-site network was relatively stable, the response latency was within an acceptable range, and the system maintained the current background compression ratio.

[0216] Engineers continued to retrieve related areas from adjacent floors, which increased the model load. In addition, some terminals were in a weak network environment in the basement. The system detected an increase in response latency and automatically further compressed the background details far away from the abnormal area, retaining only the floor skeleton and the outline of key components, thereby ensuring that the switching and zooming of the abnormal core area remained smooth.

[0217] The purpose of this step is to construct a closed-loop display system constrained by response timeliness, thereby achieving dynamic coordination between high-precision local highlighting, global context preservation, and terminal usability.

[0218] In a preferred embodiment of the present invention, the multi-source heterogeneous detection data includes engineering quality stress-strain data, concrete strength data, and settlement displacement data; the initial three-dimensional spatial model data includes building information model component data.

[0219] This embodiment provides a specific adaptation mechanism for typical engineering quality data; specifically, in order to make the aforementioned demonstration link have direct engineering applicability, this embodiment explicitly limits the multi-source heterogeneous detection data to the three most common and complementary types of data in engineering quality scenarios, namely stress-strain data, concrete strength data and settlement displacement data.

[0220] At the same time, the initial three-dimensional spatial model is limited to building information model component data; this enables the establishment of a stable mapping relationship between physical meaning, component semantics, and spatial positioning.

[0221] Stress-strain data mainly reflects the stress response of a structure under load, and is suitable for identifying problems such as sudden stress changes at beam-column joints, local stiffness anomalies, and the effects of bracing transformation.

[0222] Concrete strength data mainly reflects whether the material properties meet the design or age requirements, and is suitable for identifying problems such as insufficient early strength, uneven local pouring quality, and performance differences in repaired areas.

[0223] Settlement displacement data reflects the overall or differential deformation of the foundation and superstructure, and is suitable for identifying the impact of uneven settlement, localized deformation concentration, and construction disturbance on stability. These three types of data correspond to the stress state, material state, and geometric state, respectively, and when combined, they can provide a more complete description of engineering quality risks.

[0224] Building information model component data provides the structural semantic foundation required for the placement and display of the aforementioned data; component type, floor affiliation, adjacency relationship, cross-sectional features, and spatial coordinates can all be provided directly or indirectly by the model.

[0225] Taking the exemplary virtual verification model as an example, if strain data is bound to beam member B1, concrete strength data is bound to a certain detection surface of column member C1, and settlement data is bound to a monitoring point of foundation cap F1, then the system can simultaneously represent these three types of information in the same building information model space, and organize them into an interpretable anomaly propagation scenario based on the component relationship, rather than independent chart fragments.

[0226] Regarding the handling of anomalies, if a certain type of data is temporarily missing, such as the latest strength report not yet being uploaded, the system can still complete the preliminary anomaly display based on the other two types of data, but will mark the material evidence chain as incomplete.

[0227] If the codes of individual components in the building information model are inconsistent with the codes of the measured points on site, a one-time binding correction can be completed through the component alias table, floor section mapping, or manual verification table.

[0228] If the project is undergoing renovation or expansion or the model is outdated, resulting in significant differences between the actual components on site and the model, the system can first attach the test data to the existing model using temporary annotations, and then migrate it to the formal component object after the model is updated, thus avoiding data link interruption.

[0229] In the aforementioned super high-rise complex, the system integrates the stress and strain data of the beam-column joints of the transfer layer, the concrete strength rebound and core drilling verification results of the area, and the settlement and displacement observation data of the underground and podium connection parts, and binds them uniformly to the beam, column, wall, slab, foundation and monitoring point component objects in the building information model.

[0230] In this way, when an anomaly occurs in the southeast corner conversion layer, engineers can not only see the local stress changes, but also refer to the material properties at that location and the background of foundation deformation to form a more complete quality judgment.

[0231] The purpose of this step is to construct a data display base with clear component semantics, spatial positioning capabilities, and engineering interpretation foundation by combining typical engineering inspection data with building information models, thereby realizing the direct application of the present invention in the quality management of actual building construction projects.

[0232] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A cloud-based engineering quality inspection data visualization system, deployed on a cloud platform server, characterized in that, include: The data access module is used to acquire multi-source heterogeneous detection data and corresponding historical evolution data carrying spatial coordinates from sensor gateways, detection agency data terminals and field inspection terminals through communication networks; acquire initial three-dimensional spatial model data containing multiple spatial nodes corresponding to the multi-source heterogeneous detection data; perform format cleaning and data normalization on the multi-source heterogeneous detection data; and map the normalized data to the corresponding spatial nodes of the initial three-dimensional spatial model data based on the spatial coordinates. An anomaly assessment module is used to perform spatiotemporal feature analysis on the mapped multi-source heterogeneous detection data, calculate the anomaly weight corresponding to each multi-source heterogeneous detection data, and divide the multi-source heterogeneous detection data into local abnormal data and global normal data based on the anomaly weight. The visual routing module is used to calculate the distribution set of gravity values ​​of the local abnormal data on each spatial node in the initial three-dimensional spatial model data based on the abnormal weight of the local abnormal data, as a spatial visual gravity field, and generate visual weight parameters based on the spatial visual gravity field. The topology reconstruction module is used to reconstruct the rendering topology of the initial three-dimensional spatial model data according to the visual weight parameters, generate a local high-precision rendering model, and generate a global degraded context model according to the global normal data and the pre-configured initial context degradation compression ratio. The dynamic rendering module is used to calculate the spatial visual signal-to-noise ratio of the local high-precision rendering model and the global degraded context model in terms of the percentage of pixels displayed on the screen. Based on the spatial visual signal-to-noise ratio, the local high-precision rendering model and the global degraded context model are sent to the client for fusion display. The visual weight routing response delay during the fusion display process is obtained, and the context degraded compression ratio is dynamically adjusted based on the visual weight routing response delay.

2. The cloud-based engineering quality inspection data visualization system according to claim 1, characterized in that, The specific process by which the anomaly assessment module calculates the anomaly weight corresponding to each multi-source heterogeneous detection data includes: Time series analysis is performed on the normalized multi-source heterogeneous detection data to extract time series values ​​containing timestamps; Spatial analysis is performed on the normalized multi-source heterogeneous detection data to obtain spatial node data; The difference between the time series values ​​at adjacent time nodes is calculated as the mutation rate, and the absolute value of the difference between the spatial node data and the preset spatial reference value is calculated as the threshold deviation. Obtain the preset dimensionless spatiotemporal weight allocation coefficients with a sum of 1; Based on the weight allocation coefficient, the mutation rate and the threshold deviation are weighted and summed to obtain the anomaly weight corresponding to each multi-source heterogeneous detection data.

3. The cloud-based engineering quality inspection data visualization system according to claim 2, characterized in that, The specific process by which the anomaly assessment module divides multi-source heterogeneous detection data into locally anomalous data and globally normal data based on the anomaly weight includes: Obtain the preset anomaly detection threshold; The anomaly weight corresponding to each multi-source heterogeneous detection data is compared with the anomaly determination threshold; If the abnormal weight is higher than or equal to the abnormal judgment threshold, the corresponding multi-source heterogeneous detection data will be marked as local abnormal data. If the abnormal weight is lower than the abnormal judgment threshold, the corresponding multi-source heterogeneous detection data will be marked as global normal data.

4. The cloud-based engineering quality inspection data visualization system according to claim 3, characterized in that, The specific process by which the visual routing module calculates the spatial visual gravitational field of local anomaly data in the initial three-dimensional spatial model data includes: Obtain the spatial coordinates of the local anomaly data in the initial three-dimensional spatial model data; Calculate the spatial distance between each spatial node in the initial three-dimensional spatial model data and the local anomaly data based on the spatial coordinates; Construct a gravitational decay function with the aforementioned anomaly weight as the numerator parameter and the exponential term of the spatial distance as the denominator parameter; The gravitational value of each spatial node in the initial three-dimensional spatial model data is calculated by the gravitational decay function, and the gravitational values ​​of each spatial node are combined to form a spatial visual gravitational field. The maximum gravitational value in the spatial visual gravitational field is extracted as the gravitational extremum, and visual weight parameters are generated based on the gravitational extremum.

5. The cloud-based engineering quality inspection data visualization system according to claim 4, characterized in that, The specific process by which the topology reconstruction module performs rendering topology reconstruction on the initial 3D spatial model data based on the visual weight parameters to generate a local high-precision rendering model includes: Based on the visual weight parameters, the key rendering areas in the initial three-dimensional spatial model data are determined; Extract the evolution feature vector of the historical evolution data within the key rendering area, and the geometric feature vector of the initial three-dimensional spatial model data of the key rendering area; Based on a preset engineering risk feature mapping matrix, the evolutionary feature vector and the geometric feature vector are jointly projected onto the same risk semantic feature space for feature alignment; the cosine similarity between the aligned evolutionary feature vector and the geometric feature vector is calculated, and the cosine similarity is used as the multidimensional feature aggregation fidelity parameter. Based on the multi-dimensional feature aggregation fidelity parameters, the key rendering areas are enhanced with multiple levels of detail to generate a local high-precision rendering model.

6. The cloud-based engineering quality inspection data visualization system according to claim 5, characterized in that, The specific process by which the topology reconstruction module generates a global degradation context model based on global normal data includes: Obtain the remaining model regions from the initial 3D spatial model data, excluding the key rendering areas; Map the global normal data to the remaining model region, and calculate the initial rendering resource consumption of the remaining model region; Get the current context downgrade compression ratio; Based on the aforementioned context degradation compression ratio, the remaining model region is subjected to texture resolution reduction and rendering polygon count suppression processing to generate a global degradation context model.

7. The cloud-based engineering quality inspection data visualization system according to claim 6, characterized in that, The specific process by which the dynamic rendering module calculates the spatial visual signal-to-noise ratio between the local high-precision rendering model and the global degraded context model includes: Obtain the core pixel region corresponding to the local high-precision rendering model; Obtain the background environment pixel region corresponding to the global degradation context model; Calculate the first visual proportion of the number of pixels in the core pixel region to the total number of pixels on the screen, and calculate the second visual proportion of the number of pixels in the background environment pixel region to the total number of pixels on the screen; The quotient between the first visual weight and the second visual weight is calculated as a dynamic ratio, and the dynamic ratio is used as the spatial visual signal-to-noise ratio.

8. The cloud-based engineering quality inspection data visualization system according to claim 7, characterized in that, The dynamic rendering module fuses and displays the local high-precision rendering model and the global degraded context model based on the spatial visual signal-to-noise ratio, and dynamically adjusts the context degradation compression ratio. The specific process includes: The local high-precision rendering model and the global degradation context model are fused and rendered in the same three-dimensional coordinate system, and the visual weight routing response latency during the fusion rendering process is calculated. Obtain the preset delay threshold; The visual weighted routing response delay is compared with the delay threshold; If the visual weighted routing response latency is lower than or equal to the latency threshold, the current context degradation compression ratio is maintained, and the fusion display result is output. If the visual weighted routing response latency is higher than the latency threshold, the context degradation compression ratio is increased, and the topology reconstruction module is triggered to regenerate the global degradation context model for fusion display.

9. The cloud-based engineering quality inspection data visualization system according to claim 1, characterized in that, The multi-source heterogeneous detection data includes engineering quality stress-strain data, concrete strength data, and settlement displacement data; The initial three-dimensional spatial model data includes building information model component data.

Images