Lightweight three-dimensional modeling underground comprehensive pipe gallery intelligent data processing and monitoring system

Abstract

The invention relates to the technical field of pipe gallery data processing, and discloses a lightweight three-dimensional modeling underground comprehensive pipe gallery intelligent data processing and monitoring system comprising a space-time alignment module which generates a centimeter-level space-time coordinate system and synchronizes multi-source data; the dynamic modeling module dynamically adjusts the voxel resolution based on engineering specification parameters, performs modeling on an abnormal region, and performs sparse coding compression on a non-abnormal region; the self-organizing semantic anchor point module generates virtual anchor points at force transmission path nodes, dynamically associates data in a radius and binds a maintenance record; the anomaly deduction module deduces a risk area along a force transmission path based on a mechanical model and triggers directional scanning; the closed-loop feedback module dynamically optimizes modeling parameters and data association rules, and excessive occupation of hardware resources by traditional full-amount modeling is avoided while the real-time performance of system response is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent data processing for intelligent pipe corridors, and in particular to an intelligent data processing and monitoring system for underground integrated pipe corridors with lightweight 3D modeling. Background Art

[0002] With the acceleration of the urbanization process, underground integrated pipe corridors, as important municipal infrastructure, are playing an increasingly important role in the centralized laying and intelligent management in the fields of electric power, gas, communication, water supply and drainage, etc. However, the operation and maintenance management of integrated pipe corridors faces many challenges such as redundant data storage, low accuracy of real-time monitoring, and high consumption of computing resources, making the existing operation and maintenance systems difficult to meet the requirements of high efficiency, accuracy and low cost.

[0003] Currently, existing technical solutions mainly focus on visualization management systems based on GIS + BIM. For example, the invention patent with the publication number CN112446080A discloses "a visualization system and method for the operation and maintenance of an integrated pipe corridor based on GIS + BIM technology", which constructs a visualization system for the operation and maintenance of an integrated pipe corridor by using the GIS and BIM integration technology. This solution mainly provides operation and maintenance management support through methods such as 3D modeling, data monitoring, and remote control. The specific technical features include: establishing an overall model of the pipe corridor and surrounding facilities by using BIM 3D modeling technology; enhancing the visualization effect of the pipe corridor structure through GIS system rendering; realizing remote control and intelligent monitoring by combining artificial intelligence big data analysis; relying on Internet of Things sensors for real-time data collection, and performing data processing and updating through cloud computing. Although this solution has made certain progress in visualization operation and maintenance, there are still the following deficiencies. For example: 1. Full-scale modeling leads to high consumption of computing resources; this solution uses BIM full-scale modeling, that is, a complete 3D model of the pipe corridor is built, and monitoring is carried out on this basis. Although this method can provide an intuitive visualization effect, the data volume is huge, and the storage and computing resources are consumed extremely high. Especially during long-term operation, the cost of model update and data synchronization is relatively high, affecting the operation and maintenance efficiency. 2. Insufficient integration and spatio-temporal alignment of multi-source data: This solution relies on multiple on-site data collection terminals to monitor the state of the pipe corridor and transmits data through the Internet of Things network. However, its data sources are diverse, including fixed monitoring devices, manual mobile terminals, and passive data collection terminals. The spatio-temporal coordinate systems of various types of data are not fully unified, which may lead to mismatched data time series and large monitoring result errors. Summary of the Invention

[0004] The technical problem solved by the present invention is to provide a multi-mode power supply conversion and control system for a three-dimensional parking lot to solve the problems of storage redundancy, response lag and data fragmentation caused by the difficulty of balancing 3D modeling and real-time monitoring as mentioned in the above background art.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is as follows: An intelligent data processing and monitoring system for an underground integrated pipe gallery with lightweight 3D modeling, including a spatio-temporal alignment module, a dynamic modeling module, a self-organizing semantic anchor module, an anomaly deduction module, and a closed-loop feedback module, where:

[0006] The spatio-temporal alignment module is used to receive the multi-source sensor data stream in the pipe gallery, generate a spatio-temporal coordinate system with centimeter-level accuracy based on the predefined pipe gallery topological structure, synchronize the clocks of heterogeneous sensor data, make the timestamp error less than 20 milliseconds, and output the aligned data stream to the dynamic modeling module;

[0007] The dynamic modeling module includes: a static constraint modeling unit, which initializes the voxel resolution of each area based on the spatio-temporal coordinate system and the pipe gallery engineering safety specification parameters; a dynamic trigger modeling unit, when it detects that the pipeline displacement in the aligned data stream exceeds the preset threshold or the stress change rate exceeds the safety threshold, triggers the following operations: only improves the voxel resolution of the abnormal area to centimeter level; performs sparse coding compression on the non-abnormal area, and the compression rate is not less than 85%; outputs local modeling data to the self-organizing semantic anchor module;

[0008] The self-organizing semantic anchor module is used to receive the local modeling data, generate virtual anchors at the key nodes of the pipe gallery force transmission path, and each anchor is configured with: a three-dimensional coordinate mapped to the pipe gallery BIM model; a dynamic association radius, the size of which is determined by the pipeline type and safety parameters; based on the dynamic association radius, automatically binds the real-time monitoring data and historical maintenance records within the radius, and outputs a semantic data set to the anomaly deduction module;

[0009] The anomaly deduction module is used to receive the semantic data set, select a mechanical transmission model according to the pipe gallery structure type; deduce the abnormal diffusion range along the force transmission path, generate a coordinate set of high-risk areas and trigger the following operations: output the coordinates of the high-risk area to the dynamic modeling module to start directional scanning, and output the deduction result to the closed-loop feedback module;

[0010] The closed-loop feedback module is used to receive the deduction result, compare the deviation between the modeling data and the real-time monitoring value; when the deviation exceeds the preset threshold, dynamically adjust the voxel resolution and dynamic association radius parameters, and feedback the correction instruction to the dynamic modeling module and the self-organizing semantic anchor module.

[0011] As a further solution of the present invention, the triggering conditions of the dynamic trigger modeling unit include: the pipeline displacement exceeds the preset displacement threshold, and the preset displacement threshold is set according to the pipeline type: 3 mm for gas pipelines, 5 mm for power pipelines, and 8 mm for water supply and drainage pipelines; the stress change rate exceeds the safety threshold for 3 consecutive sampling periods, and the safety threshold is 80% of the design allowable stress; the gas concentration gradient change rate exceeds 30% of the lower explosion limit.

[0012] As a further solution of the present invention, the calculation rule of the dynamic association radius is as follows: for gas pipelines, the basic radius R = pipe diameter × 2.0, and it expands to 1.5R when leakage is detected; for power pipelines, the basic radius R = pipe diameter × 1.5, and it expands to 1.2R when the temperature exceeds 70°C; for water supply and drainage pipelines, the basic radius R = pipe diameter × 1.2, and it expands to 1.5R when the pressure fluctuation exceeds ±15% of the design value.

[0013] As a further solution of the present invention, the selection basis of the mechanical transfer model includes: for cast-in-place structures, the node shear force transfer model is adopted, and the input parameters include: the shear bearing capacity corresponding to the concrete strength grade, the ratio of stirrup spacing to diameter, and the stiffness matching coefficient of adjacent structural units; for precast structures, the contact stress diffusion model is adopted, and the input parameters include: the calibrated value of bolt pre-tightening force, the roughness grade of the contact surface, and the elastic modulus of the sealant.

[0014] As a further solution of the present invention, the benchmark point marking rule of the spatio-temporal coordinate system is as follows: a benchmark point is marked every 50 meters on the straight section, and the coordinate error does not exceed ±0.5 mm; on the elbow section, a benchmark point is marked every π / 4 radians according to the curvature radius, and the coordinate error is ±1 mm; at the center point of the pipeline intersection, a benchmark point is compulsorily marked, and the coordinate error is ±0.3 mm.

[0015] As a further solution of the present invention, the execution mode of sparse coding compression includes: for static structure areas, octree coding compression is adopted, the skeleton point cloud data is retained, and the compression ratio is not less than 90%; for dynamically changing areas, differential coding compression is adopted, the complete point cloud is retained and the time stamp sequence is marked, and the compression ratio is not less than 70%.

[0016] As a further solution of the present invention, the correction rules of the closed-loop feedback module include: if the deviation between the modeling data and the real-time monitoring value exceeds 10%, the manual review process is triggered, and the voxel resolution decision rule is updated after the review is passed; if the deviation is still greater than 5% after 3 consecutive corrections, the voxel resolution parameters are reset to the initial value, and the dynamic association radius is expanded to 1.2 times.

[0017] As a further solution of the present invention, the cooperation logic between the dynamic modeling module and the abnormal deduction module is as follows: when the abnormal deduction module outputs the influence range, the dynamic trigger modeling unit preferentially scans the high-risk areas on the force transmission path, and the scanning priority is sorted according to the stress concentration coefficient: coefficient 1.5 - 2.0: first priority, scanning resolution 0.1 mm; coefficient 1.2 - 1.5: second priority, scanning resolution 0.5 mm; after the scanning is completed, the self-organizing semantic anchor point module automatically updates the data binding rule within the association radius, and the update period does not exceed 30 seconds.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By converting industry safety specifications into dynamic voxel resolution adjustment rules and automatically embedding the design standards of utility tunnels during the modeling process, the limitation of traditional modeling that only relies on geometric accuracy and ignores engineering compliance is broken through. While the system triggers centimeter-level modeling in abnormal areas, it performs intelligent thinning on non-critical areas, realizing the simultaneous optimization of model accuracy and storage efficiency. 2. Virtual anchor points are generated based on the key nodes of the force transmission path in the utility tunnel. Through the intelligent binding of the spatial coordinates of the anchor points and the device attributes, multi-source data around the abnormal points are automatically associated. Compared with the traditional method of manually annotating the coordinate mapping table, the integrated processing of abnormal location and data traceability is realized, greatly reducing the time cost and operation risk of multi-system data retrieval. 3. Combine the mechanical transmission model according to the construction type of the utility tunnel, such as shear force transmission of in-situ structure / contact diffusion of precast structure, deduce the abnormal diffusion range along the force transmission path, and trigger high-precision modeling verification in reverse. Through the dynamic comparison of modeling data and real-time monitoring values, the voxel resolution and semantic association parameters are automatically corrected to form an active defense closed loop of "perception - deduction - verification - optimization", upgrading the traditional passive response mode to pre-event risk interception. 4. Adopt the skeleton point cloud retention and differential coding compression technology to implement high-magnification data compression for static areas while maintaining the topological integrity of the utility tunnel structure. At the same time, through the strong correlation binding of the space-time coordinate system and operation and maintenance data, it is ensured that the model state at any time point can be traced back to the original design parameters and construction records, realizing the contradiction that it is difficult to balance lightweight modeling and engineering auditing. 5. Based on edge computing chips and dynamic resolution laser devices, a collaborative processing link for lightweight modeling and real-time deduction is constructed. Through the event-triggering characteristics of pulse neural networks, computing power is preferentially allocated to high-risk area analysis. While ensuring the real-time response of the system, it avoids the excessive occupation of hardware resources by traditional full-scale modeling, providing a cost-effective solution for large-scale utility tunnel group monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0020] Figure 1 It is the core module of the intelligent data processing and monitoring system for the underground utility tunnel of the present invention and its data processing flow chart;

[0021] Figure 2 It is the working flow chart of the abnormal deduction module of the present invention;

[0022] Figure 3 It is a schematic diagram of the storage proportion of the modeling data of the present invention;

[0023] Figure 4 This is a schematic diagram of the edge computing and data storage architecture of the present invention. Detailed implementation manners

[0024] Next, the technical solutions in the embodiments of the present invention will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0025] Please refer to Figure 1 , an intelligent data processing and monitoring system for an underground integrated pipe gallery with lightweight 3D modeling, including a spatio-temporal alignment module 101, a dynamic modeling module 102, a self-organizing semantic anchor module 103, an anomaly deduction module 104, and a closed-loop feedback module 105, where: The spatio-temporal alignment module 101 is used to receive the multi-source sensor data stream in the pipe gallery, generate a spatio-temporal coordinate system with centimeter-level accuracy based on the predefined pipe gallery topological structure, synchronize the clocks of heterogeneous sensor data, make the timestamp error less than 20 milliseconds, and output the aligned data stream to the dynamic modeling module 102;

[0026] The dynamic modeling module 102 includes: a static constraint modeling unit, which initializes the voxel resolution of each region based on the spatio-temporal coordinate system and the pipe gallery engineering safety specification parameters; a dynamic trigger modeling unit, when it detects that the pipeline displacement in the aligned data stream exceeds a preset threshold or the stress change rate exceeds the safety threshold, triggers the following operations: only improves the voxel resolution of the abnormal region to the centimeter level; performs sparse coding compression on the non-abnormal region, and the compression rate is not less than 85%; outputs local modeling data to the self-organizing semantic anchor module 103;

[0027] The self-organizing semantic anchor module 103 is used to receive the local modeling data, generate virtual anchors at the key nodes of the pipe gallery force transmission path, and each anchor is configured with: three-dimensional coordinates mapped to the pipe gallery BIM model; a dynamic association radius, the size of which is determined by the pipeline type and safety parameters; based on the dynamic association radius, automatically binds the real-time monitoring data and historical maintenance records within the radius, and outputs a semantic data set to the anomaly deduction module 104;

[0028] The anomaly deduction module 104 is used to receive the semantic data set, select a mechanical transmission model according to the pipe gallery structure type; deduce the abnormal diffusion range along the force transmission path, generate a coordinate set of high-risk regions and trigger the following operations: output the coordinates of the high-risk regions to the dynamic modeling module 102 to start a directional scan, and output the deduction result to the closed-loop feedback module 105;

[0029] The closed-loop feedback module 105 is used to receive the deduction result and compare the deviation between the modeled data and the real-time monitoring value; when the deviation exceeds the preset threshold, the voxel resolution and the dynamic correlation radius parameter are dynamically adjusted, and a correction instruction is fed back to the dynamic modeling module 102 and the self-organizing semantic anchor module 103.

[0030] As a further solution of the present invention, the triggering conditions of the dynamic trigger modeling unit include: the pipeline displacement exceeds the preset displacement threshold, and the preset displacement threshold is set according to the pipeline type: 3 mm for gas pipelines, 5 mm for power pipelines, and 8 mm for water supply and drainage pipelines; the stress change rate exceeds the safety threshold for 3 consecutive sampling periods, and the safety threshold is 80% of the design allowable stress; the gas concentration gradient change rate exceeds 30% of the lower explosion limit.

[0031] As a further solution of the present invention, the calculation rule of the dynamic correlation radius is as follows: for gas pipelines: the basic radius R = pipe diameter × 2.0, and it expands to 1.5R when a leak is detected; for power pipelines: the basic radius R = pipe diameter × 1.5, and it expands to 1.2R when the temperature exceeds 70 °C; for water supply and drainage pipelines: the basic radius R = pipe diameter × 1.2, and it expands to 1.5R when the pressure fluctuation exceeds ±15% of the design value.

[0032] As a further solution of the present invention, the selection basis of the mechanical transfer model includes: the node shear force transfer model is adopted for the cast-in-place structure, and the input parameters include: the shear bearing capacity corresponding to the concrete strength grade, the ratio of stirrup spacing to diameter, and the stiffness matching coefficient of adjacent structural units; the contact stress diffusion model is adopted for the precast structure, and the input parameters include: the calibrated value of bolt pre-tightening force, the contact surface roughness grade, and the elastic modulus of the sealant.

[0033] As a further solution of the present invention, the benchmark point marking rule of the spatio-temporal coordinate system is as follows: a benchmark point is marked every 50 meters for straight sections, and the coordinate error does not exceed ±0.5 mm; for bent pipe sections, a benchmark point is marked every π / 4 radians according to the curvature radius, and the coordinate error is ±1 mm; a benchmark point is compulsorily marked at the center point of pipeline intersections, and the coordinate error is ±0.3 mm.

[0034] As a further solution of the present invention, the execution method of sparse coding compression includes: using octree coding compression for static structure areas, retaining the skeleton point cloud data, and the compression rate is not less than 90%; using differential coding compression for dynamically changing areas, retaining the complete point cloud and marking the time stamp sequence, and the compression rate is not less than 70%.

[0035] As a further solution of the present invention, the correction rules of the closed-loop feedback module 105 include: if the deviation between the modeling data and the real-time monitoring value exceeds 10%, an artificial review process is triggered, and the voxel resolution decision rule is updated after the review is passed; if the deviation is still greater than 5% after 3 consecutive corrections, the voxel resolution parameters are reset to the initial values, and the dynamic correlation radius is expanded to 1.2 times.

[0036] As a further solution of the present invention, the cooperation logic between the dynamic modeling module 102 and the abnormal deduction module 104 is: when the abnormal deduction module 104 outputs the influence range, the dynamic trigger modeling unit preferentially scans the high-risk areas on the force transmission path, and the scanning priority is sorted according to the stress concentration coefficient: coefficient 1.5 - 2.0: first-level priority, scanning resolution 0.1mm; coefficient 1.2 - 1.5: second-level priority, scanning resolution 0.5mm; after the scanning is completed, the self-organizing semantic anchor module 103 automatically updates the data binding rules within the correlation radius, and the update period does not exceed 30 seconds.

[0037] See Figure 2 FIG. is the working flow chart of the abnormal deduction module 104, which shows the working process of the abnormal deduction module 104 in an intelligent data processing and monitoring system for an underground integrated pipe gallery with lightweight three-dimensional modeling. Among them, the gas pipeline anchor point has coordinates (X1, Y1, Z1), and the correlation radius R = pipe diameter × 2.0 is set, and monitoring data such as pressure, temperature, and corrosion records are bound at the same time. When an abnormal situation is detected, the system triggers mechanical deduction and analyzes using the shear force transfer model according to the force transmission path of the cast-in-place structure. The input parameters include concrete strength and stirrup spacing, and then the diffusion range is calculated and output. The abnormal diffusion areas within the diffusion range form a high-risk area coordinate set, and are sorted according to the stress concentration coefficient > 1.5. In addition, during the deduction process, the system expands the correlation radius to 1.5R to further optimize the abnormal detection range and ensure the monitoring accuracy and data processing accuracy. Figure 3 FIG. is a schematic diagram of the storage ratio of modeling data, which shows the optimization effect of an intelligent data processing and monitoring system for an underground integrated pipe gallery with lightweight three-dimensional modeling in terms of the storage ratio of modeling data, and compares the storage ratios of key area detail retention, traditional full-scale modeling, and the compressed version of this solution. Among them, key area detail retention accounts for 50%, ensuring high-precision modeling of important pipe gallery structures and abnormal areas; traditional full-scale modeling accounts for 43%, indicating that the traditional method is prone to excessive consumption of computing resources due to large storage pressure; the compressed version of this solution only accounts for 8%. By adopting octree coding compression and differential coding compression technologies, skeleton point cloud data is retained for static structure areas, and the compression rate is not less than 90%. For dynamically changing areas, differential coding is performed using timestamp annotation methods, and the compression rate is not less than 70%. This effectively reduces the storage occupancy, makes the modeling data more lightweight, and meets the requirements of long-term monitoring of underground integrated pipe galleries. Figure 4It is a schematic diagram of the edge computing and data storage architecture, showing the edge computing and data storage architecture of an intelligent data processing and monitoring system for an underground integrated pipe gallery with lightweight 3D modeling, including the data interaction and processing flow between edge computing nodes, laser scanning devices, and data storage modules. Among them, the edge computing node adopts the SNN algorithm stack with a latency <50ms to ensure fast data processing, performs efficient calculations based on the Ascend 910B chip, and supports the retrieval of historical data to optimize data analysis capabilities; the laser scanning device has a response time <200ms and supports dynamic resolution switching, can adjust the modeling accuracy according to the real-time state of the pipe gallery, and collects data through control instructions and transmits the point cloud data stream to the edge computing node for processing; the processed modeling data is stored in the data storage module, which supports a write throughput ≥2GB / s and uses a dual-link RAID6 array for data redundancy protection to ensure the stability and reliability of long-term data storage.

[0038] Example 1: In the dynamic modeling process of this example, to ensure that the details of abnormal areas can be accurately captured and non-abnormal areas do not consume excessive storage and computing resources, the system adopts an adaptive voxel resolution adjustment strategy. For areas where stress changes or displacements exceed the threshold are detected, the system dynamically increases the voxel resolution to the centimeter level, while for non-abnormal areas, combined with the stability assessment of historical data, sparse coding compression is used to reduce the computational burden. This mechanism enables the modeling system to balance high-precision monitoring and lightweight storage. The identification of abnormal areas not only relies on single-sensor data but combines multi-sensor data fusion, and noise is excluded through statistical filtering algorithms to avoid false triggering of high-precision modeling and reduce data offsets caused by sensor errors. In terms of abnormal deduction, the system selects applicable mechanical models according to different pipe gallery structure types. For cast-in-place structures, a node shear force transfer model is adopted to ensure that factors such as concrete strength and stirrup spacing are considered during the calculation process; for precast structures, a contact stress diffusion model is adopted to ensure tightness and balanced force. To improve the deduction accuracy, the system introduces a structural stiffness matching factor to calculate the force transfer of adjacent structural units and predicts the abnormal diffusion range based on the stress gradient change to accurately lock high-risk areas and avoid ineffective scanning. The key calculation methods are as follows:

[0039]

[0040] Among them, σ transIt represents the stress gradient on the force transmission path. E1 and E2 are the elastic moduli of adjacent structural units respectively, A1 and A2 are the effective load-bearing areas, and L1 and L2 are the characteristic lengths. This calculation method can effectively identify high stress concentration areas and adjust the scanning priority accordingly to ensure the reasonable allocation of computing resources. Among them, L1 and L2 represent the characteristic lengths of adjacent structural units. For example, for a straight pipe section: in a straight pipeline structure, L1 and L2 respectively represent the effective calculation lengths between two adjacent structural units (such as two sections of pipelines, a pipeline and a support member, etc.). Specifically, L1, for example, represents the distance from the fixed support point of the pipeline to the stress concentration point, while L2 represents the length from this point to the next structural support node. These two values determine the calculation of the stress gradient during the mechanical transmission process and reflect the force balance degree of this section of the pipeline. For a bent pipe section, for example: in the bent pipe area, L1 and L2 need to be calculated in combination with the pipeline curvature. For a pipe section with a large curvature radius, L1 and L2 can be taken as the effective load-bearing distances along the pipe wall, that is, the lengths from the central axis of the pipeline to the area where the curvature changes significantly, which all belong to the extended implementation methods known to those of ordinary skill in the art.

[0041] The threshold for voxel resolution adjustment is based on industry safety specifications, such as the displacement detection limit around gas pipelines, while the calculation method of the correlation radius is based on the pipe diameter, leakage risk level, etc., rather than a fixed setting. σ in the abnormal deduction trans It is used to calculate the force transmission situation of the pipe gallery structure to ensure the rationality of the deduction model. The calculation of all variables is directly related to the technical solution, avoiding conceptual descriptions. In terms of data compression, the system adopts differential processing for different regions. The static region uses octree coding compression to reduce storage occupancy, while the dynamically changing region uses temporal difference coding to ensure information traceability while retaining key point cloud data. The compression ratio is no longer a fixed value but is adaptively adjusted in combination with the risk level. For example, the compression ratio in the low-risk region remains above 90%, while in the high-risk region, only the non-essential parts are compressed on the premise of ensuring the integrity of the modeling to ensure that the core data is not lost. The closed-loop optimization strategy of the system further enhances the real-time performance and adaptability. When there is an error between the monitoring data and the modeling data, the system will adaptively adjust the anomaly detection algorithm to reduce the false alarm rate and trigger the recalibration of the modeling parameters when the error exceeds the safety range. If the monitoring error exceeds the set threshold three times in a row, the system will reset some modeling parameters and automatically adjust the sensor calibration deviation to improve the overall detection stability. The deduction model in the high-risk region will also be updated with the changes of time and environmental factors to improve the accuracy of anomaly prediction. At the same time, it combines edge computing to optimize the data processing path to ensure rapid system response and reasonable resource utilization.

[0042] Embodiment 2: The dynamic modeling module 102 adopted by the present invention adaptively adjusts according to data such as displacement, stress, and environmental parameters collected by sensors to reduce the computational load in non-critical areas while ensuring refined modeling of abnormal areas. For example, the determination of abnormal areas is based on multi-source data fusion, not relying solely on a single sensor signal, but combining cross-comparison of data from stress monitoring, displacement measurement, and environmental sensors. The data fusion uses statistical filtering and trend analysis methods to exclude occasional noise signals and ensure the stability of trigger conditions. For example, when the displacement or stress change within three consecutive sampling periods is detected to exceed the set threshold, the system will determine it as a real anomaly and trigger centimeter-level resolution modeling. For non-abnormal areas, historical data stability analysis is used to decide whether to perform data thinning processing to optimize storage and computational efficiency. This method ensures the accuracy of dynamic voxel resolution adjustment, avoids false triggering of high-precision modeling, and improves modeling efficiency and real-time performance. The core role of the abnormal deduction module 104 is to combine the structural mechanics characteristics of the underground utility tunnel to deduce the abnormal diffusion range and give early warnings for high-risk areas. The selection of the mechanical transfer model is based on the type of utility tunnel structure, the material characteristics of adjacent components, and the structural stiffness matching situation. For cast-in-place concrete structures, the node shear force transfer model is mainly adopted, which takes into account the shear bearing capacity of concrete, the ratio of stirrup spacing to diameter, and the stiffness matching coefficient of adjacent structural units. For prefabricated assembled structures, the contact stress diffusion model is adopted, which focuses on bolt pre-tightening force, contact surface roughness, and the elastic modulus of sealant to ensure the stability of the assembled connection. In addition, the system optimizes the allocation strategy of deduction calculation resources through dynamic adjustment of the stress concentration coefficient. For areas with a relatively high stress concentration coefficient (such as 1.5 - 2.0), refined scanning is preferentially performed to ensure high-precision detection of structural weak links; for areas with a relatively low stress concentration coefficient, sparse scanning is performed using a larger voxel resolution to reduce computational pressure. This deduction mechanism combines the actual engineering requirements, enabling the system to accurately identify anomalies and efficiently utilize computational resources, thus improving the overall operation efficiency.

[0043] The closed-loop feedback module 105 plays a core role in anomaly detection and modeling optimization. To ensure the long-term stable operation of the system, the closed-loop feedback mechanism needs to compare the deviation between the monitoring data and the modeling data in real time and dynamically optimize the modeling parameters according to the deviation situation. The closed-loop feedback mechanism of the present invention is divided into two-level processing methods, such as primary correction of data deviation: when the deviation between the monitoring data and the modeling data exceeds 10%, the system will trigger an artificial review process, and after the review is passed, the voxel resolution adjustment rule will be updated to enhance the modeling accuracy; adaptive parameter reset: if the deviation is still greater than 5% after three consecutive corrections, the system will automatically reset some modeling parameters and expand the dynamic correlation radius to 1.2 times the initial value to enhance the robustness of anomaly detection. In addition, the system will perform adaptive calibration on the sensor data with large errors to reduce the false alarm rate and improve the reliability of anomaly detection. This closed-loop optimization strategy ensures that the modeling system can be dynamically adjusted according to environmental changes, avoids the decline of anomaly detection accuracy caused by long-term cumulative errors, and at the same time ensures the reasonable allocation of modeling resources, enabling the system to maintain efficient and accurate monitoring capabilities during long-term operation.

[0044] The calculation rule of the dynamic correlation radius is one of the important features of the present invention, and its rationality directly affects the accuracy of anomaly detection and data processing. For different types of pipelines, the setting of the dynamic correlation radius needs to consider the pipeline size, operating status, and safety code requirements. For example: the basic radius R of the gas pipeline is set to 2.0 times the pipe diameter and is extended to 1.5R during leak detection. This rule is based on the gas leakage diffusion model and takes into account the influence of the gas concentration gradient change to ensure that the detection range can cover the potential leakage diffusion area; the basic radius of the power pipeline is set to 1.5 times the pipe diameter and is extended to 1.2R when the temperature exceeds 70°C. This setting refers to the current-carrying capacity of the cable and the influence of thermal expansion to ensure that the overheating risk can be effectively monitored; the basic radius of the water supply and drainage pipeline is set to 1.2 times the pipe diameter and is extended to 1.5R when the pressure fluctuation exceeds ±15% of the design value. This setting combines the water hammer effect analysis to ensure that sudden pressure changes can be detected in time. The adjustment of the dynamic correlation radius uses real-time data monitoring and combines historical trend analysis to avoid misjudgment triggered by single abnormal data and improve the reliability of the system.

[0045] The spatio-temporal alignment module 101 is responsible for integrating multi-source sensor data into a unified centimeter-level spatio-temporal coordinate system. Due to the complex environment of the underground utility tunnel, sensors may be affected by factors such as temperature and humidity changes and electromagnetic interference, resulting in certain errors in the data. Therefore, the spatio-temporal alignment mechanism of the present invention adopts high-precision clock synchronization and combines a multi-sensor fusion algorithm to optimize the data alignment accuracy. The specific measures include: adopting a weighted clock synchronization mechanism, assigning different weights to sensor data with different precisions to reduce the error impact of a single data source; establishing a dynamic error compensation model, automatically adjusting the data deviation through comparison with historical data to ensure the stability of sensor data; a multi-level data alignment strategy, adopting a high-precision synchronization strategy for key monitoring data (such as displacement and stress), and adopting a lower-precision alignment strategy for environmental data (such as temperature and humidity) to reduce the calculation cost; this multi-level data fusion method ensures the accuracy of spatio-temporal alignment, provides a reliable data basis for dynamic modeling, and improves the stability and usability of the entire system, all of which belong to the extended implementation manners known to those of ordinary skill in the art.

[0046] Embodiment 3: In order to improve the detection accuracy in this embodiment, during the data fusion process, for example, a multi-level credibility screening mechanism can be adopted to make the system more robust when making decisions on abnormal situations. Specifically, before triggering high-precision modeling, the system first verifies whether an abnormal event is reliable through multi-source data fusion. The screening process includes: historical trend comparison: comparing the data of the past 5 consecutive cycles to exclude false alarms caused by single-point mutations. For example, if the stress change recorded by a certain sensor is abnormal but there is no obvious change in the historical data, the modeling is postponed and the data of the next cycle is awaited for confirmation; cross-sensor data verification: cross-comparing the data of different types of sensors. For example, if the displacement detected by a displacement sensor is abnormal while there is no leakage signal from the gas sensor, the abnormal weight of this displacement data is reduced to avoid false alarms; dynamic credibility threshold adjustment: for sensors operating for a long time, the system adaptively adjusts their credibility based on the sensor drift model. For example, newly installed sensors have a lower initial weight, while sensors that have been operating stably for more than 6 months have an increased weight to reduce false alarms;

[0047] For the fine modeling strategy of abnormal regions, this embodiment introduces a dynamic modeling trigger mechanism, such as dynamically adjusting the modeling trigger threshold: according to environmental parameters (such as temperature, humidity, ground settlement rate, etc.), the trigger threshold is dynamically adjusted. For example, in a humid environment, the stress response of some pipe materials may be enhanced, so the modeling trigger threshold can be appropriately increased to avoid over-modeling; Regional weight modeling strategy: For pipe areas with higher risks (such as gas main roads), the modeling trigger threshold is reduced, while for areas with lower risks (such as telecommunication lines), the trigger threshold is appropriately increased to reasonably allocate computing resources; In terms of closed-loop feedback adjustment, this embodiment optimizes the error correction process to ensure that the system can quickly adjust after detecting abnormalities and avoid wasting resources due to false alarms. In the original solution, the error correction strategy was based on resetting parameters when the error exceeded the standard for 3 consecutive times, but there was a lack of detailed analysis of the error source, which might lead to unnecessary adjustments. Therefore, this embodiment introduces a hierarchical correction mechanism, such as primary correction (error less than 5%): If the deviation between the modeling data and the monitoring data is within 5%, the system corrects the current data through the historical trend compensation model without directly adjusting the voxel resolution; Secondary correction (error 5%-10%): When the error is between 5% and 10%, the system will first adjust the dynamic correlation radius rather than directly increasing the modeling resolution. For example, if the data deviation of the pipeline temperature sensor is large, the system will first expand the data acquisition radius and include more sensor data for comparison; Tertiary correction (error exceeding 10%): Only when the continuous error exceeds 10% will the modeling parameters be reset and manual review be triggered to ensure that the modeling adjustment will not waste computing resources due to misjudgment.

[0048] In addition, this embodiment further optimizes the data storage and compression strategy to ensure the integrity of key data while reducing the storage overhead of unnecessary data. For example, an adaptive compression ratio is adopted: for abnormal regions, the compression ratio is reduced to ensure the integrity of details (such as the compression ratio of the gas pipeline leakage area is not less than 50%), while for long-term stable regions, the compression ratio can be increased to 90%; Key point data protection: During the storage optimization process, the system ensures that the complete point cloud of data in high-risk regions is retained instead of directly storing it with a high compression ratio. Even in low-risk regions, if a large pressure fluctuation is detected in the short term, the key point protection mechanism will still be triggered to ensure the accuracy of data backtracking, which all belong to the extended implementation methods known to those of ordinary skill in the art.

[0049] Embodiment 4: In the process of modeling the pipe gallery, the calculation of the stress gradient is crucial for the identification of abnormal regions. In its calculation formula, σ transThe stress variation between adjacent structural units is directly related to whether the force distribution is balanced; E1 and E2 represent the elastic moduli of the materials, and the specific values depend on the actual material of the pipe gallery. For example, the moduli of concrete structures and metal pipes are quite different, and they need to be distinguished during calculation; A1 and A2 refer to the effective bearing areas, which can be directly extracted during BIM modeling. For gas pipelines, the influence of the protective layer on the stress distribution needs to be additionally considered during calculation; L1 and L2 represent the characteristic lengths between units. During the calculation of straight pipe segments, the node spacing can be adopted, while for bent pipe segments, the curvature parameters need to be combined to more accurately reflect the actual stress situation. All of these belong to the extended implementation methods known to those of ordinary skill in the art.

[0050] To reduce the impact of false alarms, this embodiment supplements a more accurate false alarm screening mechanism, such as multi-sensor data comparison: if an abnormality is detected by a certain sensor, the system will automatically call the data of surrounding sensors for cross-verification. For example, if the displacement sensor shows structural deformation but the pressure sensor shows no obvious change, the system can suspend modeling and wait for the next data update; dynamic adjustment of sensor weights: for sensors that have been running for a long time, the system dynamically adjusts their weights according to historical false alarm situations. If a certain sensor has reported a large number of false alarms in the past month, the weight of its data in the decision-making will be reduced to reduce the interference with the overall judgment; trend analysis for auxiliary judgment: Abnormality triggering is not only based on single-point data, but also combines the data trends in the past several sampling periods. For example, if a certain parameter fluctuates violently in a short period of time but then returns to normal, the system will not immediately trigger high-precision modeling, but will first observe the trend changes to avoid false alarms caused by single mutations. In terms of the closed-loop feedback mechanism, when the deviation between the modeling data and the real-time monitoring value is between 5% - 10%, the dynamic correlation radius can be preferentially adjusted instead of directly increasing the voxel resolution, thereby reducing the unnecessary computational burden. If the deviation exceeds 10% for three consecutive cycles, the system will trigger manual review and recalibrate the modeling parameters to avoid instability caused by frequent adjustments. In addition, the system will continuously record the corrected data each time and optimize the subsequent modeling strategy through trend analysis. For example, if a large deviation appears in a certain area for a long time, the system will automatically adjust its basic voxel resolution to improve the accuracy of long-term monitoring. All of these belong to the extended implementation methods known to those of ordinary skill in the art.

[0051] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be encompassed by the present invention.

Claims

1. An intelligent data processing and monitoring system for an underground integrated pipe gallery with lightweight 3D modeling, characterized in that It includes a spatio-temporal alignment module, a dynamic modeling module, a self-organizing semantic anchor module, an anomaly deduction module, and a closed-loop feedback module, where: The spatio-temporal alignment module is used to receive the multi-source sensor data stream in the utility tunnel, generate a spatio-temporal coordinate system based on the predefined utility tunnel topological structure, synchronize the clocks of heterogeneous sensor data, and output the aligned data stream to the dynamic modeling module; The dynamic modeling module includes: a static constraint modeling unit, which initializes the voxel resolution of each region based on the spatio-temporal coordinate system and the utility tunnel engineering safety specification parameters; a dynamic trigger modeling unit, when it detects that the pipeline displacement in the aligned data stream exceeds the preset threshold or the stress change rate exceeds the safety threshold, triggers the following operations: only increases the voxel resolution of the abnormal region; performs sparse coding compression on the non-abnormal region, and outputs the local modeling data to the self-organizing semantic anchor module; The self-organizing semantic anchor module is used to receive the local modeling data, generate virtual anchors at the key nodes of the force transmission path in the utility tunnel, and each anchor is configured with: three-dimensional coordinates mapped to the utility tunnel BIM model, and dynamically associates the real-time monitoring data and historical maintenance records within the radius and the bound radius, and outputs the semantic data set to the anomaly deduction module; The anomaly deduction module is used to receive the semantic data set, select a mechanical transmission model according to the utility tunnel structure type, deduce the abnormal diffusion range along the force transmission path, generate a coordinate set of high-risk regions and trigger the following operations: output the coordinates of the high-risk regions to the dynamic modeling module to start directional scanning, and output the deduction result to the closed-loop feedback module; The closed-loop feedback module is used to receive the deduction result, compare the deviation between the modeling data and the real-time monitoring value; when the deviation exceeds the preset threshold, dynamically adjust the voxel resolution and the dynamic association radius parameter, and feedback the correction instruction to the dynamic modeling module and the self-organizing semantic anchor module.

2. The intelligent data processing and monitoring system for an underground integrated pipe gallery with lightweight 3D modeling according to claim 1, characterized in that The triggering conditions of the dynamic trigger modeling unit include: the pipeline displacement exceeds the preset displacement threshold, and the preset displacement threshold is set according to the pipeline type: 3mm for gas pipelines, 5mm for power pipelines, and 8mm for water supply and drainage pipelines; the stress change rate exceeds the safety threshold for 3 consecutive sampling periods, and the safety threshold is 80% of the design allowable stress; the gas concentration gradient change rate exceeds 30% of the lower explosion limit.

3. The intelligent data processing and monitoring system for the lightweight three-dimensional modeled underground utility tunnel according to claim 2, wherein The calculation rule of the dynamic association radius is as follows: for gas pipelines: the basic radius R = pipe diameter × 2.0, and it expands to 1.5R when a leak is detected; for power pipelines: the basic radius R = pipe diameter × 1.5, and it expands to 1.2R when the temperature exceeds 70°C; for water supply and drainage pipelines: the basic radius R = pipe diameter × 1.2, and it expands to 1.5R when the pressure fluctuation exceeds ±15% of the design value.

4. The intelligent data processing and monitoring system for the lightweight three-dimensional modeling of the underground integrated pipe gallery according to claim 3, wherein, The selection basis of the mechanical transmission model includes: the node shear force transmission model is adopted for the cast-in-place structure, and the input parameters include: the shear bearing capacity corresponding to the concrete strength grade, the ratio of stirrup spacing to diameter, and the stiffness matching coefficient of adjacent structural units; the contact stress diffusion model is adopted for the precast structure, and the input parameters include: the calibrated value of the bolt pre-tightening force, the contact surface roughness grade, and the elastic modulus of the sealant.

5. The intelligent data processing and monitoring system for the lightweight three-dimensional modeling of the underground integrated pipe gallery according to claim 1, characterized in that, The benchmark point marking rules for the space-time coordinate system are as follows: For straight line segments, a benchmark point is marked every 50 meters, and the coordinate error does not exceed ±0.5 mm; for bent pipe segments, benchmark points are marked every π / 4 radians according to the curvature radius, and the coordinate error is ±1 mm; at the center point of pipeline intersections, benchmark points are compulsorily marked, and the coordinate error is ±0.3 mm.

6. The intelligent data processing and monitoring system for the lightweight three-dimensional modeling of the underground integrated pipe gallery according to claim 5, characterized in that The execution methods of sparse coding compression include: using octree coding compression for static structure areas, retaining the skeleton point cloud data, and the compression ratio is not less than 90%; using differential coding compression for dynamically changing areas, retaining the complete point cloud and marking the timestamp sequence, and the compression ratio is not less than 70%.

7. The intelligent data processing and monitoring system for the lightweight three-dimensional modeling of the underground integrated pipe gallery according to claim 1, characterized in that, The correction rules of the closed-loop feedback module include: If the deviation between the modeling data and the real-time monitoring value exceeds 10%, the manual review process is triggered, and the voxel resolution decision rule is updated after the review is passed; if the deviation is still greater than 5% after 3 consecutive corrections, the voxel resolution parameters are reset to the initial values, and the dynamic correlation radius is expanded to 1.2 times.

8. The intelligent data processing and monitoring system for an underground integrated pipe gallery with lightweight 3D modeling according to claim 7, characterized in that, The cooperation logic between the dynamic modeling module and the abnormal deduction module is as follows: When the abnormal deduction module outputs the influence range, the dynamic trigger modeling unit preferentially scans the high-risk areas on the force transmission path. The scanning priority is sorted according to the stress concentration coefficient: coefficient 1.5 - 2.0: first priority, scanning resolution 0.1 mm; coefficient 1.2 - 1.5: second priority, scanning resolution 0.5 mm; after the scanning is completed, the self-organizing semantic anchor point module automatically updates the data binding rules within the correlation radius, and the update period does not exceed 30 seconds.

Images